Linkping Studies in Science and Technology

Thesis No. 953

Affordable Automation for

Airframe Assembly

Development of Key Enabling Technologies

Henrik Kihlman

Production Systems
Department of Mechanical Engineering
Linkpings Universitet
Linkping, Sweden, 2005
ISBN: 91-85299-59-6
ISSN: 0345-7524

2005 Henrik Kihlman

Production Systems
Department of Mechanical Engineering
Linkpings Universitet
SE-581 83 Linkping, Sweden
E-mail: henrik.kihlman@ikp.liu.se
Phone: +46-13-288974
Fax: +46-13-282798

Cover: The Affordable Automation Concept

Printed by: Roland Offsettryck, 2005

 Abstract
Building aircraft is a challenging field. An aircraft has a life expectancy of 40 years,
compared to just 10 years for a car. Given the vibrations of flying at close to Mach one at an
altitude of 10,000 meters, these machines must function flawlessly in a tough environment.
This demands high quality in the assembly processes. The typical part joining process in the
automotive industry is welding, whereas in the aircraft industry, assembly is made through
drilling, followed by fastening. The typical tolerances for part location in aircraft assembly, as
well as for hole drilling, is +/- 0.2 mm.

This dissertation discusses the use of industrial robots, widely used for welding and pick-and-
place operation for automotive industry, in the automation of the aircraft industry, and
specifically for the drilling of holes in the assembly process of airframe parts. The dissertation
presents how a new drilling technology called orbital drilling is incorporated with and
industrial robot. Orbital drilling reduces the cutting forces up to ten times compared to
conventional drilling using a spiral cutter.

The robot is also utilized for performing changeovers between different airframe structure
types. A novel jointed reconfigurable tooling system called Affordable Reconfigurable
Tooling (ART) is presented, which uses the robot to reconfigure flexible fixture modules. The
ART system can also be rebuilt, which means that the tool is dismantled and reused for a
completely different product family (e.g. wings, fins or fuselage sections). This is made
possible through a modular framework, i.e. not welded as with conventional tooling, but
rather jointed by screws.

Robots, originally developed for the automotive industry, have an accuracy which is ten times
less accurate than that required for aerospace applications. To help meet this limitation in the
use of robots in aircraft assembly, an additional metrology system, used in the aircraft
industry for calibrating assembly tooling, is integrated into the robot controller. The feedback
loop enables the robot to be positioned to 0.05 mm absolute accuracy. This integration is
made possible by existing embedded software packages for the robot and the metrology
system.

The processes in the system are programmed in a software package with an intuitive user
interface in a 3D-environment, normally used for the offline-programming of robots in
automotive industry. The planning is intuitive, and an approach towards a process planning
abstraction level is presented where processes are defined directly on the coordinate frames
constituting the robot trajectories and manual operations. Tolerance on accuracy requirements
are dynamically programmed in the same environment. The metrology system, working
online with the robot controller, eliminates most of the calibration work required in traditional
robot programming. Changes in the operation planning take less than a minute to run
physically with the best tolerance.
 Acknowledgements
First of all I would like to thank my Professor Mats Bjrkman who has supervised and
supported my research. I would also like to thank by co-supervisors Dr. Jonas Herbert for
great support, especially on the research methodology, Dr. Gilbert Ossbahr for good support
in the tooling section and for being a great colleague and friend over the years. I would also
like to thank all other people at the Division of Production Systems for all their support and
especially to Dr. Ekan Sundin, Dr. Mica Comstock, Kerstin Johansen, Johan stlin and Peter
Bjurstam, for fruitful research discussions and for good AfterResearch. Thanks Mica for
helping me out with the English language. Thanks Lisbeth Hgg for being there for me with
administrative issues.

Without Saab Aerostructures in Linkping this dissertation would not existed. Saab
Aerostructures is a great company and I owe many people at Saab my appreciation for great
support, especially when I worked at Saab almost full time the first year of my research. I
would especially like to thank Magnus Engstrm and Richard Lindqvist, co-authors of several
of the appended papers, for the daily contact weve had and experience from their own
industrial experience. A thank goes also to Jan Tano and Maria Weiland at Saab for
supporting me in many ways.

It has also been an honor to work at the University of Wollongong in Australia and Hawker
de Havilland for six month. I want to thank Professor Chris Cook and Dr. Steve Gower for a
generous hospitality and especially a big hug to Marta Fernandes, Stephen Van Duin and his
family for being great friends during my stay. Thanks Stephen for straighten out my
Swenglish. At Hawker de Havilland I want to thank Dr. Phil Crothers for hosting me at their
site and for helping me with this dissertation.

The first year of my research the Swedish Defense Material Administration partly funded me.
I want to thank them for a generous founding. The next three years was financed by the
European Commission, which made me work closely with the end-user aircraft builders
Airbus both in UK and Spain, BAE Systems in UK and Saab Aerostructures here in
Linkping. Throughout my time as a researcher I have also been financed by the Swedish
Foundation for Strategic Research through the Programme for Production Engineering
Education and Research (PROPER) and ProViking. The support is gratefully acknowledged.

I would like to thank the people at Novator AB in Stockholm, especially Ingvar Eriksson and
Mark Ennis who are both co-author in one of my appended papers. Thanks Ingvar and Mark
for the support of information and expertise on orbital drilling.

I also want to thank Mahboob Alam from Boeing Long Beach California for very interesting
discussions weve had at the SAE conferences and for the support you gave me in this
dissertation. The same appreciation goes to Leo Muys from Stork Fokker in the Netherlands
for a great openness in showing me your research and for the help you gave me in this
dissertation.

Finally I would like to thank my family back home in Lidkping for all their support over the
years. Especially my mom who tells me not to work so much!
LINKPING, MAY 2005
Henrik Kihlman
 Appended Publications
During the course of the project resulting in this dissertation, the author presented findings,
some of which were the result of cooperation with others, at various international conferences
and/or in academic journals. Eight of these publications, which are appended in full, are
listed bellow.

Appended Publications

Paper I Kihlman, H., (2001) Reconfigurable Tooling for Airframe Assembly A

state-of-the-art Review of the Related Literature and a Short Presentation of a
New Tooling Concept, CIRP 1st International Conference on Reconfigurable
Tooling, May 21-22

Paper II Herbertsson, J., Kihlman, H. and Engstrm, M., (2002) "Reconfigurable

Aircraft Assembly Using industrial robots and new tooling to meet future
production scenarios ", 33rd International Symposium on Robotics in
Stockholm, October 8-10

Paper III H. Kihlman and M. Engstrm, "Affordable Reconfigurable Tooling", SAE

2002 Transactions Journal of Aerospace, ISBN: 0-7680-1285-6

Paper IV Kihlman, H., Eriksson, I. and Ennis, M., (2002) "Robotic Orbital Drilling of
Structures for Aerospace Applications ", SAE Aerospace Automated
Fastening Conference & Exposition, October. 1-3

Paper V Kihlman, H. and Loser, R., (2003) "6DOF Metrology-integrated Robot

Control", SAE 2002 Transactions Journal of Aerospace, ISBN Number: 0-
7680-1448-4, pg 398

Paper VI Kihlman, H., Sunnanbo, A., Loser, R., Von Arb, K., Cooke, A., (2004)
"Metrology-integrated Industrial Robots Calibration, Implementation and
Testing", 35th International Symposium on Robotics, Paris-Nord Villepinte,
France, March 23-26

Paper VII Kihlman, H., Ossbahr, G., Engstrm, M., Anderson, J., (2004) "Low-cost
Automation for Aircraft Assembly", 2004 Aerospace Manufacturing and
Automated Fastening Conference & Exhibition, September 20-23, Sheraton
West Port Hotel, St. Louis, Missouri

Paper VIII Lindqvist, R, Kihlman, H.,(2004) "Orbital Drilling - Implementation and

Evaluation", 2004 Aerospace Manufacturing and Automated Fastening
Conference & Exhibition, September 20-23, Sheraton West Port Hotel, St.
Louis, Missouri
 Contents
THESIS STRUCTURE...............................................................................................................1

PART I: BACKGROUND AND STATE-OF-THE-ART..........................................................3

1 Introduction ........................................................................................................................ 5
1.1 Background ............................................................................................................ 5
1.1.1 Aircraft Market................................................................................................... 5
1.1.2 Aircraft Assembly .............................................................................................. 9
1.1.3 Aircraft Assembly Automation ........................................................................ 10
1.2 Objectives............................................................................................................. 13
1.3 Delimitations ........................................................................................................ 14
1.4 Research Methodology......................................................................................... 15
1.4.1 Introduction ...................................................................................................... 15
1.4.2 Research Design............................................................................................... 16
1.5 Dissertation Structure........................................................................................... 20
2 Aircraft Production........................................................................................................... 23
2.1 The Airframe Product Families and Assembly Levels ........................................ 23
2.1.1 Product families................................................................................................ 23
2.1.2 Assembly levels................................................................................................ 23
2.2 Aircraft Production Process ................................................................................. 26
2.3 Aircraft Materials ................................................................................................. 27
2.3.1 Metallic Materials ............................................................................................ 27
2.3.2 Non-metallic materials ..................................................................................... 27
2.4 The Aircraft Assembly Process............................................................................ 28
2.5 Machines for Aircraft Automation ....................................................................... 31
2.5.1 Large Scale Automation................................................................................... 31
2.5.2 Parallel Kinematic Machines ........................................................................... 33
2.5.3 Serial Kinematic Industrial Robots .................................................................. 35
2.6 Industrial Robots for Aircraft Automation........................................................... 37
2.6.1 Accuracy........................................................................................................... 37
2.6.2 Dynamic Loads ................................................................................................ 39
2.6.3 Work Envelop .................................................................................................. 41
2.6.4 End-effectors .................................................................................................... 41
2.6.5 Final Remarks Using Robots for Airframe Assembly ..................................... 42

PART II: AIRFRAME ASSEMBLY FOR AFFORDABLE AUTOMATION....................... 45

3 Robotic Orbital Drilling ................................................................................................... 47
3.1 Conventional Drilling........................................................................................... 47
3.1.1 Reaming ........................................................................................................... 48
3.1.2 Plug Gauges...................................................................................................... 48
3.1.3 Countersinking ................................................................................................. 49
3.1.4 Deburring ......................................................................................................... 49
3.2 Orbital Drilling..................................................................................................... 49
3.2.1 Fatigue analysis ................................................................................................ 51
3.2.2 Cost modeling .................................................................................................. 51
3.3 Force Measurements ............................................................................................ 52
3.3.1 Experimental Equipment.................................................................................. 52
3.3.2 Forces in Conventional Drilling ....................................................................... 53
3.3.3 Forces in Orbital Drilling ................................................................................. 55
3.3.4 Force Comparison ............................................................................................ 57

i
 3.4 Robot Orbital Drilling .......................................................................................... 57
3.4.1 The Preliminary Test Bed Using The PODU................................................... 57
3.4.2 The Final Test Bed Using the CNC Orbital Drilling Unit ............................... 61
3.5 Final Remarks on Robotic Orbital Drilling.......................................................... 63
4 Affordable Reconfigurable Tooling ................................................................................. 65
4.1 Aircraft Tooling Theory ....................................................................................... 65
4.1.1 The Generic Tooling Model............................................................................. 65
4.2 Existing Tooling Solutions................................................................................... 67
4.2.1 Conventional Tooling....................................................................................... 67
4.2.2 Modular Tooling .............................................................................................. 67
4.2.3 CNC-controlled Tooling .................................................................................. 68
4.2.4 The Need for New Tooling Technology .......................................................... 70
4.3 Affordable Reconfigurable Tooling ..................................................................... 71
4.3.1 Modular Framework......................................................................................... 71
4.3.2 Dynamic Modules ............................................................................................ 73
4.3.3 ART Pick-ups................................................................................................... 78
4.4 Physical Demonstrator ......................................................................................... 78
4.5 ART Database ...................................................................................................... 82
4.6 Additional Virtual Case-studies Using The ART System.................................... 83
4.7 Economical Aspects of the ART System ............................................................. 84
4.8 Similar Research Attempts to Robot-manipulated Tooling ................................. 85

PART III: METROLOGY-INTEGRATED ROBOTICS FOR AFFORDABLE

AUTOMATION....................................................................................................................... 87
5 Metrology for Maintained Accuracy................................................................................ 89
5.1 Metrology Possible for Robot-integration............................................................ 90
5.1.1 Laser Trackers .................................................................................................. 90
5.1.2 Photogrametry Systems.................................................................................... 92
5.1.3 Indoor Positioning System ............................................................................... 95
5.2 Metrology Systems Used In This Research ......................................................... 97
5.2.1 The LTD800..................................................................................................... 97
5.2.2 The T-Cam and T-Probe .................................................................................. 98
5.2.3 The 6DOF Metrology Reflector ....................................................................... 98
5.3 Similar Concepts to This Research .................................................................... 100
6 System Integration.......................................................................................................... 103
6.1 Embedded System Control................................................................................. 103
6.1.1 WebWare For Robot Integration................................................................. 103
6.1.2 Generic Robot Programming Using WebWare.............................................. 105
6.1.3 emScon For Tracker Integration ................................................................. 107
6.2 System Topology................................................................................................ 109
6.2.1 Network Topology ......................................................................................... 109
6.2.2 System Overview ........................................................................................... 110
6.2.3 Robot Chuck End-effector ............................................................................. 111
6.3 Example of Probing, Docking and Reconfiguration .......................................... 113
6.4 The Integration Software.................................................................................... 115
6.4.1 Control Panel.................................................................................................. 115
6.4.2 The Operator Panel......................................................................................... 116
6.4.3 The Probe Menu ............................................................................................. 117
6.5 Metrology Feedback Controller Calculations .................................................... 118
6.5.1 Translational Error ......................................................................................... 118
6.5.2 Orientation Error ............................................................................................ 119
ii
 6.5.3 Termination Criteria....................................................................................... 119
7 System Setup Calibration ............................................................................................... 121
7.1 6D-Reflector Calibration.................................................................................... 121
7.2 Robot Chuck Calibration.................................................................................... 122
7.3 Robot Base and TCP0 ........................................................................................ 124
7.4 End-effector calibration...................................................................................... 125
7.4.1 Online calibration of End-effectors................................................................ 125
7.4.2 Offline calibration of End-effectors ............................................................... 126
7.5 Probing of Dynamic Modules ............................................................................ 127
7.5.1 Probing Using a Non-calibrated 6D-Reflector............................................... 127
7.5.2 Probing Dynamic Modules using a calibrated 6D-Reflector ......................... 129

PART IV: OPERATION PLANNING FOR AFFORDABLE AUTOMATION .................. 131

8 Introduction to Robot Programming and Simulation..................................................... 133
8.1.1 Abstraction Levels in Robot Programming.................................................... 136
8.1.2 CATIA and DELMIA .................................................................................... 139
9 Operation Planning Methodologies................................................................................ 141
9.1 Designing an ART system.................................................................................. 141
9.2 A Process-oriented Programming Approach...................................................... 146
9.2.1 Manual Tasks ................................................................................................. 148
9.2.2 Automatic Tasks............................................................................................. 149
9.2.3 The Transition between Virtual and Physical World ..................................... 151
9.2.4 Final remarks on the process-oriented programming approach ..................... 152
9.3 Definition of Kinematics in Dynamic Simulation.............................................. 152

PART V: DICUSSION AND CONCLUSION...................................................................... 155

10 Future Improvements ..................................................................................................... 157
10.1 Introduction ........................................................................................................ 157
10.2 Modular Framework........................................................................................... 157
10.3 Dynamic Modules .............................................................................................. 158
10.3.1 Ball-joints ....................................................................................................... 158
10.3.2 Un-docking a DM........................................................................................... 158
10.3.3 Bearings.......................................................................................................... 158
10.4 Pick-ups.............................................................................................................. 159
10.5 System Integration.............................................................................................. 159
10.6 Operation Planning............................................................................................. 159
11 Discussion and Conclusions........................................................................................... 161
11.1 Critical Review of the Research Questions, Objectives and Aim...................... 161
11.2 Discussion of the Research Results.................................................................... 166
11.3 Continuing Research .......................................................................................... 167
12 References ...................................................................................................................... 169
Appendices ............................................................................................................................. 177
Appendix A: Adjusting a Hexapod to High Accuracy........................................................... 178
Appendix B: The Hydraulic Diagram .................................................................................... 179
Appendix C: A Short Summary of Transform Equations ...................................................... 180
Appendix D: The Force Cone Concept .................................................................................. 182

Appended Publications........................................................................................................... 183

iii
iv
 Abbreviations
ADFAST Automation for Drilling, Fastening, Assembly, Systems Integration, and
tooling, a FP5, EU sponsored research project
CAD Computer Aided Design
CAM Computer Aided Manufacture
CFC Carbon Reinforced Composites
Clamp A holding device in an assembly tool. Typically a handle is used
DOF A degree of freedom is when the motion of a joint can be described by one
independent variable.
End- The tool or sensor that is mounted at the distal end of a robot
effector
Fixture An assembly tool that holds parts during assembly
Jig A assembly tool that holds parts during assembly and for guiding cutting tools
PKM Parallel Kinematic Machines
Prismatic A linear joint found in Cartesian robots
Revolute A type of rotary joint found in anthropomorphic robots
TCP Tool Center Point
Stack term used to describe the total thickness of various layers of wing skin,
internal flanges and reinforcing plates sandwiched together
Swage The act of pulling and compressing the sleeve of an aerospace locking bolt to
compress an assembly joint (acts like a nut on a bolt)
Swarf Coolant, chips and residue resulting from cutting operations

v
vi
 Thesis Structure
The following table enables readers to individually tailor their reading depending on interest
and time available. The table presents the five parts of the dissertation and the major issues
addressed in each, and guides the reader to their location (see also table of contents).

Part I: Background and State-of-the-Art

1. Introduction Addresses the motivation for the significance of this research;

discusses the aircraft market trends, the aircraft assembly process,
and aircraft automation; and details the structure of this dissertation.

2. Aircraft Describes the essence of aircraft product and production processes;

Production discusses the machines typical for todays airframe assembly
automation; and presents the requirements from an airframe
standpoint on using robots for automation.

Part II: Airframe Assembly for Affordable Automation

3. Robotic Orbital Reviews the Conventional Drilling method; reviews its problem
Drilling areas; describes the Orbital drilling method; presents a method for
comparing the two drilling methods; discusses their conceptual
differences; and presents the Robotic Orbital Drilling Approach.

4. Affordable Describes the essence of aircraft assembly tooling; presents the

Reconfigurable existing solutions; points out a gap of tooling technology; and
Tooling presents the new concept of Affordable Reconfigurable Tooling.

Part III: Metrology-integrated Robotics for Affordable Automation

5. Metrology for Describes the motivation for using metrology; discusses existing
Maintained solutions for metrology-guiding robots; presents the system used in
Accuracy this research; and positions the authors solution to other existing
solutions.

6. System Presents the included systems for the metrology-integrated solution;

Integration discusses the programming technique; describes the system
topology; presents the integration program; and presents the
measurement feedback control loop.

7. System Setup Explains when calibration is required; and presents the calibration
Calibration calculation procedures.

Part IV: Operation Planning for Affordable Automation

8. Introduction to Introduces robot programming and geometry simulation; addresses

Robot abstraction levels of robot programming; and briefly presents the
Programming simulation platform used in this research.
and Simulation

1
9. Operation Describes the operation planning methodology for the tooling design
Planning and process planning for the tooling approach in this research;
Methodologies discusses the process-programming approach; and presents the
definition of kinematics in the tooling components.

Part V: Discussion and Conclusion

10. Future Critically reviews what is required to undertake before an industrial

Improvements implementation of the Affordable Automaton approach.

11. Discussions and Summarizes the results and critically reviews progress made towards
Conclusions the research questions; highlights the contribution of this research;
and discusses future research directions.

2
 PART I
BACKGROUND AND
STATE-OF-THE-ART

Part I introduces the field of aircraft manufacturing, beginning with an

introduction to the aerospace market, in order to set the context for better
understanding the need for developing new production technology in the area.
First, a short introduction to airframe assembly and airframe assembly
automation is presented. Following this, the structure, design and method of the
dissertation are presented.

3
4
 Introduction

1 Introduction

This Chapter introduces the reader to the area of aircraft manufacturing, from a top-level
aircraft market perspective down to specific aspects of aircraft automation. Following this
background, the research problem, purpose, scope, and significance are presented.

1.1 Background
The background to the research presented in this dissertation was formed in close dialog with
several end-users from the aerospace industry. This section will give the reader an
understanding of what market factors are driving the requirements for new production
technology in this area. Starting from the top-level market demands down to an understanding
of the aircraft assembly process and automation, this section will point out the gap of
technology within aircraft automation that has been the point of departure for this research.

1.1.1 Aircraft Market

The generation of research questions in production technology, from the top level marketing
needs, is a long path. This section provides a short overview of the aircraft market situation,
which in turn gives a holistic view for the reader to further quantify the relevance for the
research question presented at the end of this chapter.

Conquest of the skies has liberated us from the bonds imposed by geography, terrain and
water. Air routes can be thought of as the highways of the global economy, transporting
people and goods over vast distances at great speed, and the aviation business has thrived over
the last century, facilitating both business and leisure opportunities. This trend will without a
doubt continue to increase. Market predictions for the aerospace industry pointed towards
increased demand for air travel and steady growth in the market for aerospace products and
services until before the terrorist attack on the World Trade Center in New York September
11th 2001. As can be seen in the graph in Figure 1.1, this event caused a major impact on the
industry with air traffic in many sectors being eroded and numerous operators going out of
business or filing for bankruptcy. This event resulted in delays or cancellations of orders for
all major airframe suppliers. A Similar, but less obvious event was the SARS epidemic during
early 2003; thus event is also evident in Figure 1.1, where a small dip can be identified.

Figure 1.1: Delivered civil aircraft in the world.

Source: Teal Group and Saab Aerostructures

5
Introduction

Terrorist attacks and epidemics have made the aerospace market a fragile one which is now
highly sensitive to world events. Although recovery has been undertaken, further
developments, particularly in relation to global security, have had negative impact on the
industry.

The global market for large passenger jets is a duopoly in which two major players (Boeing
and Airbus) are locked in a strategic battle to achieve commercial dominance (Pritchard,
2002). According to Niosi and Zhegu (2002), the four major civil aircraft prime contractors
are Airbus and Boeing for planes over 100 seats, and Bombardier and Embraer for regional
jets. Later in this section, the market outlooks from Boeing (Boeings Current Market
Outlook, 2004) and Airbus (Airbus Global Market Forecast, 2004) will be used as a reference
to help estimate the state of the aircraft market for the next 20 years. A forecast by the aircraft
manufacturer Airbus states that the trend for air travel growth will average 5.5% per year over
the twenty-year period, i.e. 2004-2023. Boeing has made a similar prediction for air travel to
increase, but at an average rate of 5.2% annually.

Although many of the traditional national operators have encountered difficulties, 2002 saw a
significant increase in passenger numbers for budget airlines such as Easyjet and Ryanair - a
trend reflected in the major orders for the industry during that same year. Budget airlines have
thus far primarily penetrated domestic markets and shorter-haul international routes, hence the
need for single-aisle aircraft in the size category 100 to 200 seats. Airbus predicts that single-
aisle aircraft will constitute two-thirds of all new deliveries by 2023, while Boeing predicts
single-aisle jets will constitute three-quarters of the fleet by the same year. Boeing forecasts
that the world commercial aircraft fleet will more than double during the period 2004-2023,
reaching almost 35,000 airplanes. This would mean that for the next twenty years, 25,000 new
passenger and cargo jets will have to be built at a cost of 2 trillion USD. In the Airbus vision
from 2004 to 2023, the estimate was slightly more modest, with a forecast for 17,300 new
passenger and freighter aircraft to be built at a cost of 1.9 trillion USD.

The commercial aircraft sector is a large industrial field. For the US, this market accounts for
8% of the nations total industrial export (Pritchard, 2002). By 2000, there were 1,220,000
aerospace employees in the world, of which 49% were based in the United States, 35% in the
European Union, 7.5% in Canada, 2.7% in Japan and 5.7% in the rest of the world (Niosi and
Zhegu, 2002). In order to exploit this market, it is essential to improve the technological base
in order to increase the share of this rapidly expanding, high-technology market.

Even though it stands clear that the aircraft market is expanding, the competition will
undoubtedly harden in a globalised market, as increased globalization has led to fewer
manufacturers. One example is the U.S. aerospace industry, which in 1980 had over seventy
aerospace suppliers, but by 2004 just five prime contractors. It is also less common today that
aerospace suppliers build their own complete products in-house. The aerospace industry is
no different from other market segments when it comes to outsourcing, whereby another
partner performs activities that are outside the core technology/business of the company, and
who may be better suited to perform these activities. The typical aerospace company that
previously built complete aircraft in-house have now become one company in a network of
suppliers, where each airframe supplier is responsible for different parts of the development
and manufacturing of the final aircraft product. In a globalised world, airframe suppliers are
more likely to become specialized in a particular field of manufacturing rather than rely on
having generalized knowledge. Airbus went beyond outsourcing and formed a partnership of
companies; as a result, the company now employs 52,000 people of over 80 nationalities, and
where the major partners are located at France, Germany, Spain and the United Kingdom

6
 Introduction

(Airbus, 2005). Airbus is more of an example of an extended enterprise than an outsourcer. In

an extended enterprise, the companies involved are on more equal terms than when compared
to the relationship with contracted manufacturers (Browne and Zhang, 1999). Companies like
Airbus, Boeing, Bombardier and Embraer are so-called Original Equipment Manufacturers
(OEMs), and can be thought of as occupying the apex of a supply chain pyramid as shown in
Figure 1.2.

Tier one prime aerospace

OEM manufacturers

Tier two aerospace suppliers

Tier three aerospace component

suppliers

Figure 1.2: The three tiers of the aerospace industry

Niosi and Zhegu (2002) explained the three tier pyramid as follows:

On the top of the pyramid one finds the airframe assemblers (prime contractors or
OEMs) such as Bombardier, Embraer, Airbus or Boeing. These companies design
aircraft, prospect markets and order subassemblies to the second tier. In this
second stage we find manufacturers of propulsion systems such as General
Electric, Pratt & Whitney, or Rolls Royce, as well as producers of on-board
avionics, airframe structures and subassemblies such as landing gears, ailerons
etc. The tier 2 group of producers is concentrated at the global level with a
handful of firms dominating each segment. At the base of the pyramid one finds
tier 3, manufacturers of parts and components that are assembled by tier 2 (and
sometimes by tier 1) firms. The third component of the pyramid is largely
dispersed, with hundreds of firms competing among themselves.

The higher a firm climbs in the pyramid, the more prestigious, but also the more knowledge
about the product is required, including design, development and production. At the same
time, tier 2 suppliers share the risk capital on new investment with their tier 1 counterparts.
Many previous OEM firms are no longer building their own aircraft, yet still want to stay high
in the pyramid in order to maintain their knowledge and competence in developing and
building airplanes. Subcontractor firms on the lower tier of the pyramid still build airframe
subassemblies, but these projects are called Build-to-print projects, and hence include no
part of the development.

A trend today is the industrial offset agreement, which is organized as a production-sharing

agreement where the seller transfers portions of the manufacturing to the buyer, something
which is often required to penetrate new markets (MacPherson and Pritchard, 2002). For
example, in order to sell Boeing 747s to Air China, at least part of the final product must be
manufactured or assembled inside China itself. However, MacPherson (2003) claims that no
exporter would enter into an offset agreement unless there were significant competitive
pressures to do so. Pritchard (2002) discusses how delivering aircraft production capability
(technology and skills) to potential competitors is similar to transferring production skills
from one nation to another, and how this will make countries such as Russia and China

7
Introduction

serious players in the global market in the next 10 years. A company like Boeing has followed
a system integration strategy, which essentially means that it buys components from overseas,
and assembles aircraft at home (Pritchard, 2002).

Working as one subcontractor in a network of companies and making a part of a product, the
deadline within each sub-step of a products development and manufacturing becomes
increasingly important. This is significant because, in the end, many sub-parts from different
partners are brought together in the final assembly (sometimes called final body join), making
each deadline increasingly important. Furthermore, this place increased pressure on each
supplier to have high delivery performance, low cost and high quality, especially because
suddenly each supplier is now exchangeable. In other words, it is easier to change suppliers in
a company network model than terminate a division in a company that manufactures the
complete product in-house.

Aerospace industry, compared to most other industry sectors, has low product volumes. For
example, a car model is manufactured in the range of hundreds of thousands, while aircraft
seldom reach more than one thousand. This means from a manufacturing point of view that
the aerospace industry would require a higher need for flexibility in order to reuse production
equipment. It is always the number of products being built that pays for the investment of
production equipment; hence, to lower the cost of production equipment in the aerospace
industry, reuse would be required. An extreme scenario in this aspect is to build just one
aircraft. The next generation of fighter aircraft, the unmanned vehicle, obviously does not
require pilots to be trained.

The Swedish Air Force currently has an order placed for 200 of the Saab Gripen fighter
aircraft. This number of aircraft is required in part for the training of pilots. The next
generation of fighter aircraft is likely to constitute Unmanned Aerial Combat Vehicles
(UACVs). These kinds of aircraft have no pilot that needs training; hence, the aircraft is only
a platform for carrying new technology and weapons. In this scenario, perhaps the number of
aircraft necessary is in the range of ten, just enough to create a platform to evaluate systems
for the next generation of fighter aircraft. This would certainly intensify the need to reuse
production equipment in order to make it more affordable. In addition, in the event of war, the
number of aircraft would have to be increased at short notice; this means that the production
equipment must be quick and easy to replicate. The production equipment would now be
required to be ramped up with short lead times. As a new technology, UAVs will require
many changes to its original design in their first generations; hence the production system
would be required to quickly adapt to design changes without long lead times.

If European or U.S. aerospace builders (OEMs and suppliers) are to stay on the upper tiers of
the pyramid of Figure 1.2, and thus maintaining their development and production skills, then
they must stay competitive by making their production technology as competent as the
products they are building. The work with traditional drilling of holes and riveting, i.e. built-
to-print, has no economical incentive to stay in Europe or the U.S., especially since this can
be done at a tenth of the cost in low-wage countries. In keeping the production of airframes in
Europe and U.S., new flexible production equipment is critical in order to stay competitive
and compete with low-wage countries on a global market. This becomes especially important
as OEMs have shown signs to reduce the number of players in the higher level of the tier
pyramid.

8
 Introduction

1.1.2 Aircraft Assembly

Thus far in this Chapter, an overview view of the aerospace market trends and possible
scenarios has been presented. In this section, an introduction to aircraft assembly is given,
which will bring the reader closer to the objectives of this dissertation. Aircraft manufacturing
covers the whole area of part design, NC-machining, and manufacturing of carbon fiber
material, whereas aircraft assembly concerns the assembly of the airframe from parts to
substructures to a final airframe.

Product prerequisites for airframe assembly are challenging. The most important demand is to
have high fatigue resistance, as the consequences of fatigue in aerospace applications can be
devastating. Flying at 10,000 meters requires the hull to be airtight, and since the wings and
sometimes the aircraft fuselage are filled with fuel, it must also be fuel-tight. Aircraft can be
considered as complex vehicles. In addition, all the hydraulics and electronics that distribute
energy and signals between systems place requirements on the airframe design as well. To
give a concrete comparison, a typical automobile is composed of approximately 20,000
components, while a Boeing 777 is composed of 4 million. In general, a spot weld gun in car
assembly is positioned within +/- 1.2 mm (Axelsson, 2002); while a drilling machine in
aircraft assembly requires positioning within +/- 0.2 mm and sometimes with higher
tolerance. Finally, while a car typically has a life span of ten years, it is not unusual for
aircraft to be built to last forty.

Aircraft assembly tooling is employed for holding the aircraft parts in space during assembly.
Tooling is divided into two different principal groups: fixtures and jigs. Fixtures position and
hold parts during assembly, whereas jigs not only position and hold parts during assembly,
but are also used to guide cutting tools. As aircraft parts have been attached to the fixture
clamps, the assembly process is engaged. The assembly process is basically carried out by
drilling holes followed by fastening. The joining elements used in the aircraft assemblies are
rivets. In detail a typical airframe assembly process can be structured according to:

1. Pre-assembly
2. Drilling
3. Temporary fastening
4. Deburring
5. Sealant application
6. Fastening

These six production operations are presented in detail in Section 2.4. The airframe assembly
processes are used when joining parts in subassemblies that eventually become products in
different product families. As presented further in Section 2.1.1, examples of product families
include wings, tail sections, and fuselage parts. The most common materials in aircraft
manufacturing are aluminum, carbon fiber reinforced plastics (more generally called carbon
fiber composites), and titanium. In Section 2.3, a short summary of all common material used
in an airframe is presented. The materials are stacked up in two to four layers (called stacks)
and then joined together.

Given the challenging nature of manufacturing and creating tooling for a complex product
the aircraft the challenge is to stay a competitive actor in a globalised market, and to do so
by building aircraft with high delivery performance (products delivered on due dates), low
cost and with high quality.

9
Introduction

1.1.3 Aircraft Assembly Automation

The cost for the assembly of an aircraft consumes up to 40% of the total airframe
manufacturing cost (Bullen, 1997). The manufacturing process of Airbus wings requires the
drilling of over 40 million holes in aircraft structures per annum. The majority of these holes,
approximately 80% of the total, were drilled manually (Latger et al., 2002). A Boeing 747-
400 contains 6 million parts, of which half are fasteners, and which involves a tremendous
amount of drilling and fastening operations. The manual drilling is accomplished by highly-
skilled and trained operators using expensive tools and equipment in order to retain the
necessarily high levels of quality expected in the industry. The use of computer-controlled
machines also has its benefits; according to Sarh (2002), these machines outperform humans
by a factor of 10 to 20 times. Despite this, the technology within the aerospace industry in
assembly has remained relatively unchanged since the 1970s. Conservatism in the industry
and the ever increasing price of technology dictates that processes have evolved rather than
taken major step changes.

Another industrial field that is used throughout this dissertation to make comparison to the
aerospace is the automotive industry. This industry sector has many repetitive operations
which motivate automation. Aerospace applications, however, have many different operations
which are far less repetitive when compared to the automotive industry. For example, a
typical scenario in the aerospace industry would be the manufacturing of 1,000 aircraft or
less. Assume for this example that the number of holes to be drilled in just one structure is
1,500. These 1,500 holes would require 1,500,000 operations. In this situation, the number of
repetitive operations is only 1,000 (the size of the total product volume). In a typical car
manufacturing scenario, however, the product volume can reach 200,000 cars. That is
equivalent to 200,000 repetitive operations for the spot welding process, the prevalent
assembly method in the automotive industry. The author means here that the high number of
repetitive operations is one of the reasons the car assembly is a highly automated
manufacturing field.

Industrial robots normally used for spot welding in the automotive industry are designed to
conduct repetitive operations; hence, they have good repetitive accuracy. The absolute
accuracy in industrial robots, however, is 10-15 times less accurate than compared to the
repetitive accuracy, meaning that the challenge to automate an aerospace application is far
greater than that in the automobile industry.

Automation is not only focused on replacing labor in production; it also provides improved
environmental conditions for employees working in the assembly of aircraft and other large
and complex structures. This is achieved by improving the current level of safety for
operators through the removal of some of the more difficult of the manual assembly activities.
Presently, the work often requires the handling of large subassemblies where visibility and
positioning of the complete component is not always possible for a single operator, and
therefore increases the level of risk for the operator or the product. Better working conditions
for the operator through reduced contact with vibrating tools can also be achieved. This is
particularly important within the current confines of enclosed structures. For example, there is
still a vast amount of manual riveting being carried out in the aerospace industry. This
contributes to hand arm vibration syndrome (HAVS) and also noise and subsequent hearing
problems. Furthermore, there is often close contact with sealants and adhesives and, with the
growth of the use of carbon fiber reinforced plastics (CFRP), carbon dust. Additionally, the
effect of low access, in terms of restricted movement for the operator, provides additional
ergonomic issues such as bending, falling and cuts. Where automation is used, these risks are
reduced or removed.
10
 Introduction

In the aerospace industry, automation has existed for many years. To ensure accuracy,
stiffness in dynamic processes and to cover the product envelope, typical automation
machines for aerospace are large and dedicated machines. Such machines, examples of which
are presented in Section 2.5.1, are extremely reliable and suited for their job. The downside to
this approach, however, is that these machines are expensive; they become dedicated to the
narrow scope they were designed for. Changing to a new product that the machine was not
originally designed for has proven to be difficult when trying to reuse dedicated machinery. It
is common that these machines become obsolete and simply stand still like monuments. The
author discusses Monumental Automation as follows:

Monuments are tools, equipment or machines designed for a specific purpose in

automation. Once that purpose changes, these systems are very difficult and
expensive to reconfigure to the new purpose, and therefore often become obsolete.
Monumental Automation prevents easy reconfiguration of a production line upon
new customer demands.

The aerospace industry, where large dedicated machines have been the common method for
automation, is now striving to reduce costs and shorten lead times. These machines have
enabling high accuracy in dynamic operations, such as drilling and the assembly of high-
quality products. These machines, however, are expensive and lack flexibility.

The automotive industry has had great influence in the design of industrial robots. Today,
industrial robots are mass-produced at a rate of 100,000 robots per year, and with an
operational stock of 880,000 (World Robotics, 2004). For many years and to, industrial robots
have been used in aerospace industry in some specific application. There have, however, been
several unsuccessful attempts as well. Some industrialists are of the opinion that industrial
robots cannot be used for drilling since they not designed for aerospace applications.
Considering the fact that the dedicated machinery in Monumental Automation costs
sometimes up to 10,000,000 Euro, if a robot costing around 45,000 Euro could be used to
perform the same operations, the cost savings would be tremendous. Granted, large machines
manage large work envelopes, so the price of just one robot is a bit naive of a figure. But the
cost for the extra equipment necessary to reach the full work envelope is far from reaching the
cost for dedicated machinery such as that mentioned above. Beyer (1999) summarized the
problems involved using large machines to cover large volumes with very high positional
accuracy as follows:

Ambient temperature
Subsistence of foundations
Misalignment of axes
Non-linearity of axes
Non-rectangularity of axes

The normal method is to try controlling this environment and calibrating the machines, which
requires significant initial and recurring costs and down times.

At Hawker de Havilland, an Australian aerospace company, robots and light weight

equipment have been used to automate airframe assembly. Crothers et al. (2004) calls this
approach Lightweight Automation. This approach does not only include robots, but also the
entire approach of using relatively inexpensive production equipment for aircraft
manufacturing. The building blocks in Lightweight Automation are industrial robots,
reconfigurability, short lead times, metrology-integration and off-the-shelf components. As
11
Introduction

Dr. Crothers (2005) explained:

Instead of the traditional aerospace manufacturing approach, having steel to

give accuracy, in light weight automation, computers and integrated systems
provide the high accuracy

In the Lightweight Automation project undertaken by Hawker de Havilland (HdH) a

metrology system was used to position a drilling machine attached to an industrial robot with
high accuracy.

Another agile machine for automation is the parallel kinematic robot concept originally from
NEOS Robotics called the Tricept. Compared to a conventional robot system, a Tricept robot
improves:

Stiffness
Repeatability
Positional accuracy
Portability
Operation to CAD data

Boeing implemented a project incorporating a metrology system to the Tricept robot the TI2
concept. The TI2 concept got its name from using a Tricept robot, a 3D camera from Imetric
and the IGRIP offline programming system. The TI2 system guides the Tricept robot to
ascertain higher positional accuracy in a machining application instead of using expensive
mechanical systems (Beyer, 1999; CUMULI, 2000; Fayad et al., 2002; Whinnem, 2000).
Compared to standard industrial robots the Tricept robot has the advantage of being stiffer in
its design. The downside, however, is the rather small work envelope and the complex
structure of a Tricept that hinder access to complex structures. The Tricept robot is presented
more thoroughly in Section 2.5.2.

So far in this section, it might seem that 100% automation might be the only salvation for
lowering costs in aircraft assembly, but this is not really the case. It is important to realize that
automation has not an end in itself. As Bullen (1999) stated:

Automation should be applied only where it makes sense and where several of the
following will result: higher output, better quality, reduced scrap and rework
improvements in workspace safety and fewer required people.

Bullen further stated that an automation level of 70-80% would be a credible accomplishment.
People are an integral and important part of the assembly process; in some cases, they can
perform some operations better than automation. One thing, however, is clear: in order to
remain competitive, airframe manufacturers must reduce manufacturing costs while
maintaining product quality and schedule performance. Bullen (1999) presents ideas on how
to gradually implement low cost automation, such as:

Assess whether tooling is needed at all, and if so, how the process could be automated
to simplify or reduce requirements for tooling
Soft tooling (data) would always be preferable to physical (hard) dedicated tooling,
which is unavoidable in some cases. Soft tooling data would be delivered in a
seamless manner straight from Engineering directly to the cutter point on the factory
floor
12
 Introduction

Flexible tooling, capable of being reconfigured to suit multiple assembly operations;

where setup phase of operations is non-value added, tools will be required to have
quick set-up capabilities

Finally, it is fair to say that having the facts about the prerequisites of aircraft manufacturing
and market trends, there are strong indications that automation for airframe assembly requires
flexibility to be cost-effective.

1.2 Objectives
As was indicated in the introduction, there is great potential for using new technology in the
aerospace industry in order to make the jump up to the next generation of aircraft automation.
Technology that for many years has been used for automation in the automotive, white-goods
and electronic industries has the potential for being used also in aerospace industry. Industrial
robots are one key technology which is needed to implement a more cost-effective automation
in an industry where new ideas for bringing technology forth have been exposed to a dragging
of traditions and myths. The objective for this research is:

To develop a concept for Affordable Automation, including development of key enabling

technologies, which will reduce lead time and cost in airframe assembly.

Research Design
The objective is formulated rather broadly, but will nevertheless maintain a focus on
aerospace production technologies. In this dissertation, this objective will be reached by
discussing the answers this researchs results to five research questions. These research
questions are treated with coherence in this dissertation, and follow the following sequence:

1. How can new drilling technologies enable the use of industrial robots for aircraft
assembly automation?

The first research question will address the difficult task of using industrial robots in drilling
automation, but also investigate new drilling technologies that make drilling automation using
industrial robots easier. This first question is answered primarily in Chapter 3. Assuming the
first research question is answered, there will be industrial robots to perform drilling
automation. The question posed here is how to use these robots to perform changeover in
reconfigurable tooling. Given this, the second question is:

2. How can new assembly tooling be developed that uses an industrial robot to
perform changeover between products within a product family, and with a short
changeover time?

Products within a product family have a geometrical commonality and the geometrical range
between products within a product family is small. The answer to the second question will
end up in having a reconfigurable assembly tool for airframe assembly, where flexible tooling
modules, in the range of decimeters, enable changeover within a product family. This enables
products with a limited geometrical difference to be built in the same tooling. But what if
there is a completely different product design that is to be built, hence a product from another
product family, which would require a greater geometrical change of the tool? This brings us
to the third research question:

3. How can the assembly tooling in research question two be developed to enable
changeover between product families?

13
Introduction

Both the second and the third research questions will be answered in Chapter 4. Using
industrial robots that have been developed for the automotive industry, but are here used to
perform drilling and changeovers in tooling and rebuilding and reconfiguring tooling to form
a completely different layout, will still leave a problem open: in the previous subsection, it
was mentioned that robots have ten times less accuracy than required for aircraft
manufacturing. Thus, the fourth question is:

4. How can the absolute accuracy of industrial robots be increased to maintain the
demands placed on them by aircraft applications?

The fourth research question will be answered in Chapter 5. Earlier in this chapter, the fact
that aircraft manufacturing is a low-volume product scenario was mentioned. This means that
robots have to be reprogrammed for new products in order to maintain a high level of
utilization in the production equipment. This leads us to the last research question for this
dissertation:

5. How can new operation planning methods simplify programming of the

production system presented in this research?

The answer to the fifth research question will present methods for simplifying the strategies in
designing the assembly cell and equipment, as well as minimize the lead time from simulated
and programmed robots to the execution of programs on the workshop floor. This research
question is mainly answered in Chapter 6.

These five research questions are considered in several research subprojects, and they also
have a close relation to the appended papers. As a quick guide to which research papers are
related to the research questions stated above, the following table is provided (Table 1.1).

Table 1.1: The relationship between the research questions and the appended papers
Research
Appended Papers
Question
1 Paper IV, VIII
2 Paper I, Paper III, VII
3 Paper III, Paper VII
4 Paper V
5 VII

Research questions two and three primarily address the research described in the authors
Licentiate thesis (Kihlman, 2002). In the research for the Licentiate thesis was presented a
tooling concept that had been brought about in the research team. The research team is
presented in Section 1.4.2. This tooling concept was further built, evaluated and is presented
in this dissertation. In addition, the authors Licentiate thesis also pointed out certain key
enabling technologies required to make the conceptual idea work in reality. These key
enabling technologies were later tested and evaluated, and are presented in this dissertation.

1.3 Delimitations
The approach of the research presented in this dissertation has taken the starting point of
when the product has been designed. This dissertation does not consider how the product is
designed to suite the assembly process, e.g. Design For Assembly (DFA). There is extensive
research being carried out in DFA. One example of successful aerospace related research
using methods to reduce tooling components is JAM (Jiggles-Aerospace Manufacturing),
14
 Introduction

under development by Cranfield University (Burley et al., 1999; Naing et al., 2000). The
research presented in this dissertation is in no way in conflict with that approach, instead the
author considers the concept presented in this dissertation and JAM to complement each
other. Key-characteristics in JAM to avoid complex tooling for complex products and also to
simplify the next stage of the build is important prerequisites for Affordable Automation. In
addition, a technology like JAM will likely use some kind of reconfigurable tool to hold the
extremities of the product. It is the hope that this research can benefit to the JAM approach.
From that statement, the research presented in this dissertation does not concern design
restrictions on the product. The product requirements are used as input to the Affordable
Automation approach presented in this dissertation.

In Chapter 2 the robot concept Tricept from NEOS Robotics is presented. The research
presented in this dissertation might benefit from the stiffness characteristics in that robot
concept. The author of this dissertation has not physically evaluated the Tricept concept in
relation to the serial kinematic robot concept chosen for this research. The author has,
however, presented a summary of the Tricept robot in Section 2.5.2, and made a qualitative
evaluation in the end of Section 2.6.5 of the two robot concept.

The discussion in Section 1.4.1 presents the different research areas that are covered in this
dissertation. It is not possible to penetrate to the depth in each of these areas as would be
expected by an expert reader in just one of the topics. The author has chosen to keep this
dissertation with a balanced depth in each area.

1.4 Research Methodology

1.4.1 Introduction
Leedy (1997) presents a number of criteria for planning and designing a research project. In
planning a research project, certain features common to all research should serve as
guidelines. All research is tested by certain criteria that must be built into the research design
in the planning stage. These standards are summarized as follows:

Universality. The research project should be such that it could be carried out by any
competent person other than yourself.
Replication. The research should be possible to replicate. This means that any
competent person should be able to repeat what one researcher has already done
instead of having to reinvent the wheel.
Control. An experiment should be repeated under the identical condition and in the
identical way in which it was first carried out. This shows consistency within the
research design.
Measurement. The data should be susceptible to measurement. This is easy in physical
sciences but more challenging to quantify in humanistic and social research.

The author has implicitly used the four aspects of planning presented above throughout the
research as a general methodology. The research presented in this dissertation is more of a
practical research method than theoretical. Section 1.4.2 discusses the research design in
detail.

15
Introduction

Metrology
Drilling
Airframe-
assembly Operation-
Robotics planning
Tooling

Figure 1.3: The technology areas covered in this dissertation

The research in this dissertation covers five major areas: robotics, drilling, tooling, metrology
and operation planning, and all with a focus on airframe assembly, as illustrated in Figure 1.3.
A single researcher working in several different fields has a limited about of time and
resources to penetrate each field in the depth that someone only interested in one field would
prefer. This approach has a disadvantage, however: there is a risk that the depth in each field
risks becoming more shallow than if only one field was penetrated. As is described in Section
1.4.2, the author has utilized students in subprojects as icebreakers in the investigation of
the different technology fields. The author has used such projects to increase the depth of
knowledge that could be attained, much more so than if the author had done everything alone.
In many cases, further research studies have not been initiated for reasons of relevance in
helping to answer the overall research question. Once a subproject has been determined to
show potential in an area important to this research, the author has assumed the responsibility
and continued the work carried out by the student or students. One example was in the area of
simulation using RobCad, which occurred during the first years of this research, and
simulation using DELMIA during the latter years of the research. In this area, the author had
great help from students in subprojects where the applicability of functionality to the research
questions was evaluated. If the functionality in the simulation system or the method in the
approach was deemed to be sufficiently related to the researchs objective, they were adopted
by the author, further studied and applied to the complete system that grew bigger over the
years.

1.4.2 Research Design

The author has worked in close relation with industry in the development of the technology
presented in this dissertation. The author has also worked in a network with the five other
major technical universities in Sweden, traveling to conferences throughout the world. A great
deal of the authors research has been conducted in cooperation with a diverse group of
students, both Swedes and those from other countries. The authors research design is
summarized in Figure 1.4.

16
 Introduction

2000 2001 2002 2003 2004 2005 2006 2007 2008

NFFP2+ FlexAA

ADFAST FINABOX

CRC- IMST

Academic Relations
- PROPER
- ProViking
- Conferences
- Licentiate thesis
- Collaborative research

Industrial Relations Publications

- Co-development The Authors - SAE
- Applications Research - ISR
- Keynote speeches - RMS
- Expertise

Student Project #1
Student Project #...
Student Project #27

Figure 1.4: Research design

Student Projects
One important contribution to the result presented in this dissertation is from students that
have been deeply involved in the project almost continuously from day one. These
subprojects have been formulated and supervised by the author. The student projects have
essentially taken three forms:

Small projects, as part of courses conducted by the department

Project courses, with work in 100% project format
10 credit Bachelor theses and 20 credit Master theses projects

Students in the projects represented both Sweden and a host of other countries. These projects
have all made a great contribution to the results presented in this dissertation, and are
therefore referenced throughout its pages and included in the reference chapter as well
(Chapter 12).

The author implemented a student project management method called Academic Quarters that
included two toll gates throughout the project. The first Academic Quarter was undertaken
two weeks after the student had started the project, while the second was held after 8 weeks
for a 10 week project or 12 weeks for a 20 week project. With this method of project
management, by the presentation day, the last element in the student project, the students
exhibited familiarity and comfort when presenting the project before a group. In hindsight,

17
Introduction

involving students in this research has been a successful key factor in reaching the results that
have been achieved thus far.

Publications
Most of the conference and journal publications in this research were presented in cooperation
with the Society of Automotive Engineering (SAE). At the SAE Aerospace Congress,
attendees shared their knowledge about aircraft design, manufacturing and safety. The author
has presented several papers at the Aerospace Automated Fastening Conference & Exhibition,
which focused more on the manufacturing of subassemblies and final assembly technologies
and processes. The author has also participated twice in the International Symposium on
Robotics, organized by the International Federation of Robotics (IFR). The IFT conference
gave the author an opportunity to display this research from a more technical point of view
than the more applications-focused conferences attended (i.e. SAE). The author also
participated in a CIRP conference, essentially organized by the University of Michigan, called
the International Conference on Reconfigurable Manufacturing (RMS). The RMS paradigm
can be described as one working towards handling rapid change in the system configuration,
its machines and controls in order to quickly adjust production capacity and functionality in
response to market changes. Two journal papers where accepted in the Journal of Aerospace
(Paper III and V).

Industrial Relations
The author has thus far cooperated in two industrial projects for this research. The first
project, the NFFP2+ project, was made in collaboration with Saab Aerostructures. In this
project, a broad study was conducted through interviews with personnel at many different
departments, work training at the workshop floor and several project reports. The project was
part of the National Aerospace Technology Research Program, funded by The Swedish
Defense Material Administration. The project was a pre-study for the applicability of flexible
tooling for airframe assembly, and its purpose was to evaluate the affordability of different
assembly tooling concepts with different levels of flexibility.

The NFFP2+ project was accomplished through an investigation of the state-of-the-art in

order to discover new tooling solutions that could influence flexibility. The concepts where
compared to a reference technology, that is to say a method most commonly used at Saab
Aerostructures conventional tooling. Several concepts were generated and an economical
evaluation was performed. The author worked actively in that project to first map available
tooling technologies that influences flexibility, where flexible tooling was compared with
conventional tooling as a reference. 75% of the authors daily work the first year was spent at
Saab Aerostructures.

The NFFP2+ project overlapped with a more extensive EU-project called ADFAST
(Automation for Drilling, Fastening, Assembly, Systems Integration and Tooling). The
ADFAST project, which focused on the development of new assembly methods and
equipment for aircraft development and manufacturing, was a three-year project funded by the
European Union under the Framework V growth program. The project covered the assembly
process of aircraft manufacturing for both civil and military applications, and was divided into
three work packages: the first covered drilling and riveting applications; the second, new
novel-jointed assembly tooling; and the third, system integration. The consortium working on
the project included, in addition to Linkping University, five aerospace end-users (Airbus
UK, Airbus Espa a, Alenia, and BAE SYSTEMS) and four suppliers (Hyde, Novator AB,
mTorres and Leica). Se Figure 1.5:

18
 Introduction

AEROSPACE MANUFACTURERS

AIRBUS-UK
AIRBUS-SPAIN
SAAB AB RESEARCH ORGANISATIONS
ALENIA
BAE SYSTEMS
LINKPINGS -
UNIVERSITET
SUPPLIERS

HYDE GROUPE
MTORRES
NOVATOR
LEICA

Figure 1.5: The ADFAST consortium

Linkping University, where the author has conducted his research, had both an objective part
in evaluating technology and methods, and a role of bringing forth new blue sky ideas. In a
large project such as ADFAST, it is difficult to decide who has done what. In most cases,
ideas were generated from discussion between all partners involved in the different work
packages. Essentially, however, the role of the aircraft manufacturers was to define
specifications and provide orientation to the research organizations involved in the project,
and subsequently to test the prototypes or concepts developed.

Linkping University provided the special technical and theoretical robotic skills, associated
human and materials resources and developments that will be needed. The key activities were
to:

study other concepts of machine for automatic assembly in a close collaboration

with partners involved in the parallel tests program,
conduct the sample tests,
facilitating dissemination of results within the European Union in addition to the
end users,

The tooling, drilling and metrology suppliers, in the ADFAST project, prepared the output of
the project for exploitation. More specifically their roles were:

integrate the expected end-effectors specifications,

design and made the best end-effectors selected,
participate to the test program,
develop the equipment for full market implementation

Also shown in Figure 1.4 is two ProViking funded projects: FlexAA and Finabox, which are
denoted by the dashed lines. The FlexAA project continues the research from the ADFAST
project with a focus on faster sensor feedback characteristics on the robot, cheaper metrology
systems and the industrialization of the tooling concept that is presented in Chapter 4. This
project is not discussed in this dissertation. The Finabox project is a future project that is not
discussed in this dissertation. Briefly, the Finabox project has the objectives of develop a
factory in a box that can be shipped on a truck and ramped up in production within 24 hours,
performing manufacturing in various industrial sectors of Swedish industry.

19
Introduction

Academic Relations
The author has also been a member of the PROPER program throughout this research.
PROPER, which stands for Programme for Production Engineering Education and
Research, gave the author the opportunity to get to know doctoral students at most of the
major Swedish technical universities. The PROPER program was funded by the Swedish
Foundation for Strategic Research. PROPER also gave the author the opportunity to meet
many companies and institutes of production engineering research in Sweden and Germany.

In the later stages of the research the author has been a member of the ProViking program,
funded by the Swedish Foundation for Strategic Research. ProViking gave the author the
opportunity to meet other researchers with a broader scope in manufacturing within the areas
of product development, manufacturing, product support and maintenance in life-cycle
perspectives.

The author also spent six months in Australia at the University of Wollongong (UoW). UoW
participated with a similar project to ADFAST, but in Australia it was called the CRC IMST:
Project SP1.7: Intelligent Aerospace Assembly System, which in itself was funded by the
Cooperative Research Centre for Intelligent Manufacturing Systems & Technologies Limited
(CRC-IMST). The CRC-IMST is a company established by five universities, thirteen
manufacturing companies and the CSIRO-MS&T. It is one of over 70 CRCs established with
the aid of Commonwealth Government grants and matching contributions from the partners.
The author was generously funded to travel to and stay at the UoW. The aircraft
manufacturing company Hawker de Havilland (HdH) was a partner in that project. This
adventure was undertaken just before the writing of this dissertation, giving the author a great
opportunity to evaluate and reflect upon earlier research results and at the same time gain
more knowledge of similar research. However, the author will present only certain research
that was developed at the time in Australia, which is not conflicting with confidentiality
agreements. There will be some references to technology and methods from the UoW/HdH
project that are relevant to this dissertation.

1.5 Dissertation Structure

Before concluding this introductory chapter with abbreviations, the following chapters of the
dissertation are described:

Chapter 2 gives the reader an introduction to the field of aircraft assembly and the machines
that are commonly used in automating the assembly of airframes. Chapter 2 will also discuss
the requirements placed on the automated assembly of airframes given its aerospace
applications.

Chapter 3 will present the first of four key enabling technologies the Orbital drilling method
which creates only small forces in the drilling process, an important aspect in robotic
drilling. Orbital drilling also enables one-way assembly, eliminating part dismantling for
deburring.

The fourth chapter presents the second key technology Affordable Reconfigurable Tooling
(ART). ART enables easy changeover between products within a product family with short
lead times, but also changeover between product families (rebuild) with longer lead time.

The fifth, sixth and seventh chapters cover Metrology-integrated Robotics, which is a key-
method to inducing the capability of robots to handle the challenging process prerequisites for
airframe assembly.
20
 Introduction

Chapter eight and nine present the last key enabling technology the Operation planning
approach to this research - that provides a seamless transfer from the virtual world to the
physical world.

Chapter 10 is the first chapter in the discussion section of the dissertation, discussing some of
the future improvements required to make the essence of this research industrialized.

Chapter 11 is moving towards the final conclusions of this dissertation and Chapter 12 is the
reference chapter.

The last parts of the dissertation are including the appendix sections and after that are the
appended papers.

21
Introduction

22
 Aircraft Production

2 Aircraft Production

In Chapter 2, the reader is provided with an introduction to aircraft production in order to

understand what is an airframe, what materials constitute the airframe, what are the
assembly levels and methods utilized in airframe production and what are the machine-
related alternatives for building the airframe.

2.1 The Airframe Product Families and Assembly Levels

Product families concern different subparts of an aircraft structure. Each product family
involves several assembly steps, or levels of assembly.
Vertical
stabilizer
Wing

Fuselage

Horizontal
stabilizer

Figure 2.1: Theoretical fuselage layout and components

2.1.1 Product families

Most aircraft are built in different sections that are subsequently put together in a final
assembly stage of the assembly process. In an earlier phase of this process, the sections are
divided into subsections, see Figure 2.1. One subsection is a panel structure that constitutes
the outer skin of the airplane; together, this structure can be viewed as a distinct product
family. Another product family could consist of a wing structure or a body structure, and so
on. The definition of product families in this dissertation is not based on a way to logically
define the different parts of the aircraft, but instead as a way to - from an assembly tooling
point of view - define different substructures and the complexity they cause for the assembly
system. For example, a panel structure is a less complex structure to build than a closed-box
wing structure. In addition, two panel structures are included in a wing structure.
Furthermore, the product families are divided into different assembly levels, which is a way to
classify product families from a machine access point of view.

2.1.2 Assembly levels

Aircraft parts have been presented previously in this dissertation as high quality products,
which in turn lead to high demands on the assembly processes. The complexity of the product
is also due to the many parts included in an airframe. Due to the high demands on accuracy,
all parts need to be located correctly using assembly tooling. To further discuss tooling and
processes in this dissertation, the author has divided the products assembly stages into the
following different levels:
23
Aircraft Production

1. First level: Stringers and stiffeners are assembled to panels

2. Second level: Inner structures are assembled and combined with the first panel
3. Third level: Inner structures and the first panel are further combined with the second
panel that closes the structure
4. Fourth level: The substructures are combined into the complete fuselage section that
finally constitutes the complete fuselage.

The assembly levels are further presented individually with examples.

First assembly level

The first assembly level (Level One Assemblies) is the open
panel structure as shown in Figure 2.2a. This structure is what
constitutes the skin of the aircraft. The assembly steps for the Panel
first level concerns joining stringers and stiffeners to panels.
Tooling used to hold the products need only access from one Stringers

direction, making them well-suited for CNC-controlled tooling

methods as presented in Section 4.2.3. The stringers and
stiffeners require a tremendous number of holes to be drilled and
fasteners to be installed. Therefore, the Level One Assemblies Figure 2.2: The first
are well-suited for C-frame machines, which are presented in assembly level
further detail in Section 2.5.1.

Second assembly level

In this dissertation, a Level Two Assembly is defined as
Spar
requiring access from at least two directions. Level Two
Assembly concerns the joining of the aircraft skeleton structures,
containing spars and ribs, and applying the first panel to the
substructure, as seen in Figure 2.3. Some of the Level Two Ribs

Assemblies are quite open structures, enabling easy machine and

tooling access, and thus making them well suited for C-frame
drill and fastening machines similar to those discussed for the
Level One Assembly structures. Whereas Level One Assemblies
are part of the Level Two Assemblies, the Level Two Spar

Assemblies are produced in lower throughput compared to the Figure 2.3: The
Level One Assemblies, making them more suited for use in second assembly level
industrial robots (presented in Section 2.5.3) or parallel
kinematic robots (presented in Section 2.5.2).

24
 Aircraft Production

Third assembly level

At the third assembly level, the open box structures are closed, see Figure 2.4. For example, a
skeleton structure from the Level Two Assemblies containing
one panel structure from the Level One Assemblies are now
added to an additional panel which closes the box. Normally,
this is performed by first attaching the additional panel and
drilling the first holes with manual methods, and then
performing the rest of the drilling automatically. When all of
the drilling is complete, the structure is removed and
deburred. At this stage, it is important that there are no dust
or waste inside the structure before putting the panel structure Second panel

back again for fastening. Chapter 3 in this dissertation Figure 2.4: The third
presents a new drilling method, where dismantling the panel assembly level
after drilling is not necessary.

Fourth assembly level

The fourth assembly level is defined as the final assembly of
the airframe, see Figure 2.5. In this stage, the different
substructures, such as wing boxes, the rear and front fuselage,
midsections etc. are combined together to form a final
fuselage. At this stage the subsections are craned into jack
stands that can be manipulated for positioning the subsections
relative to each other. It is common to use a ring-riveter to
access the fuselages surroundings. In a fully automated Figure 2.5: The
system a drilling robot is on the inside moving synchronously fourth assembly level
with a robot on the outside, such as was the case with the
ARAS robot, presented in Section 2.5.1. Sahr (2002) postulated this assembly level should
mostly be done by humans due to the lack of innovation and development efforts to find and
develop flexible, adaptable and affordable assembly system alternatives. There are new daring
technologies to make these operations in spaces where a human or normal robot cannot reach;
one such technology is the Snake-arm Robot (Buckingham and Graham, 2003;
Buckingham, 2003).

25
Aircraft Production

2.2 Aircraft Production Process

This subsection presents an example of a typical production process flow for a product, from
customer order to the finished aircraft. The production process flow as presented in Figure 2.6
is simplified, but has the purpose of positioning the airframe assembly in the total production
process flow.

Order from Machining Airframe

customer assembly

Sheet metal
Order Part assembly System
planning integration
Composite
manufacturing
Final painting

Sub-suppliers Testing and

evaluation

Delivery to
customer

Figure 2.6: Production process flow from customer order to delivery

The production process flow from customer needs to final delivery can be described, as one
example of an aircraft product for Saab Aerostructures, as follows:

A customer places an order that triggers a production plan, which is made by the order
planner, which in turn triggers different manufacturing workshops, both internal and external.
Each workshop delivers partly finished parts to either the part assembly workshop, or directly
to the airframe assembly workshop. One typical example is that the composite workshop can
deliver complete substructures, such as fins, stabilizers, aileron structures etc. In the airframe
assembly workshop, subassemblies are assembled together into subgroups. After this, these
subgroups are assembled into main groups, such as rear and front fuselage, wings,
midsections etc. The final stage in the airframe assembly workshop is to combine and
assemble main groups with the complete aircraft. Some of the hydraulic pipes are installed
already in the airframe assembly workshop as well. The completion of the aircraft, i.e.
installing systems, is done in the final aircraft integration workshop. After all additional
systems, such as avionic systems have been installed, the airplane is painted to customer
requirements. Before delivering the aircraft, it is tested and evaluated to customer
requirements at the final testing facility.

26
 Aircraft Production

2.3 Aircraft Materials

The most common structure materials used in an airframe can be divided into two main
material groups: metallic materials and non-metallic materials. Figure 2.7 shows the material
composition for airframes:
Miscellaneous 1%
Titanium 5%
Steel 9%

Composites 10%

Aluminum 75%

Figure 2.7: Material composition of airplane materials (Source: Boeing Commercial

Airplanes Group, 2005)

2.3.1 Metallic Materials

The three most common metallic materials used in airframes are:

Aluminum
Stainless steel
Titanium

Aluminum is the most widely used airplane material, as it is highly formable and with good
machining characteristics, and is used throughout the airframe. However, Aluminum, together
with low-alloy steels, are the two groups of airplane materials most susceptible to corrosion.
Aluminum alloys are coated with a corrosion-inhibiting primer to protect them from
corrosion. Stainless steel has a higher strength property and thus is used where the demands
for strength are higher, but the downside is that it is a denser and heavier material. Typical
areas for stainless steel are applications where stress and fatigue causes high demands on
material properties. Typical example areas are locking devises, hinge lines, landing gear units
etc. Stainless steel parts are cadmium plated and primed if they are attached to aluminum or
alloy steel parts. This is to prevent the stainless steel from galvanically corroding the
aluminum or alloy steel parts. One difficulty is that aluminum and stainless steel have
different electrical potential, where one material becomes an offer anode, hence requiring an
interface sealant. Titanium is typically used in applications with high strength requirements,
and especially if heat is applied. Typical areas for titanium use are in the engine exhausts and
the engine bay. Titanium is lighter than stainless steel, making it a preference for that reason
on larger critical structures. The downside with titanium is that it is a very hard material,
which reduces tool life on cutter tools in drilling. Titanium is also an expensive material.

2.3.2 Non-metallic materials

Non-metallic materials are divided into two groups:

Carbon fiber reinforced composite

Kevlar

27
Aircraft Production

Carbon fiber reinforced composite is referred to in this dissertation as simply carbon fiber or
CFC. CFC is a laminate of carbon fiber impregnated with epoxy. CFC has a very good weight
and mechanical strength ratio. CFC can be used for primary structures, as skins, access panels
doors etc., as well as secondary structures such as ribs, spars, stringers, beams etc. CFC has
good machining characteristics, but the cutters become worn out. Manufacturing CFC is
considered an expensive process due to the material cost, but also due to production tooling,
time-consuming material handing and the hardening cycle. Another challenging is related to
CFC machining is that delamination can occur. CFC itself is corrosion resistant, but stacked
together with aluminum can cause galvanic corrosion in attached aluminum structures.

Kevlar is more durable but is also more cumbersome in machining due to fringes. Kevlar is
typically used for primary structures with armor properties.

Design drivers for material selection traditionally come in the following prioritized order:

Weight
Strength
Cost

2.4 The Aircraft Assembly Process

This section summarizes typical operations for the assembly process of aircraft structures.
Further explanation of technologies and processes in airframe assembly builds upon this
section, which will give the reader understanding of the processes involved. It is assumed here
that there are no changeovers needed to be done. This is assuming the fixtures are already
manufactured and a new assembly is to be built.

Pre-assembly
The pre-assembly step is about putting parts to be assembled into the assembly fixture. The
fixtures hold the parts throughout the assembly process. In most applications parts are placed
manually into the fixtures. This may, however, sometimes require cranes where parts can
weight several tons. A thorough presentation of fixtures is found in Chapter 4 on Flexible
Tooling.

Drilling
Drilling is done either manually using handheld machines, or through the use of automated
drilling systems, where the sequence of operations is numerically stored in a controller.
Automated drilling systems are more common for simpler geometries that have good access
(e.g. fuselages and skin panels). The most common method is the use of linear guided cutters.
In most cases, conventional drilling does not alone reach the high demands for quality, and
for that reason reaming is a common process following drilling. More advanced cutters have a
reaming section integrated into the cutter, so when drilling is done reaming is followed as part
of the feeding cycle. The setting of speeds and feed of the drilling machines are dependent on
the material to be drilled, as well as on the machines performance. Normally a stack of
different materials are drilled with constant cutting parameters, but this may cause problems.
A rule of thumb is that carbon fiber materials have higher cutting speeds than aluminum,
while titanium has lower cutting speeds, but with much higher torques. To facilitate the
material prerequisites, cutting parameters are preferred to change during the drilling process.
The typical cutter materials are high speed steel, carbide, or polycrystalline diamond.

28
 Aircraft Production

In drilling, hole quality requirements can be defined into two categories: dimensional
requirements and damage-related requirements (Eriksson, 2005). Moreover:

Dimensional requirements (Figure 2.8 shows some of the symbols):

Hole diameter tolerance (dcyl)

Countersink diameter, angle and depth tolerances (dc and )
Roundness (ovality)
Straightness (angularity)( )

Figure 2.8: Graphical view of the dimensional hole requirements

Damage-related requirements (Figure 2.9 shows some of the symbols):

Surface finish (hole and countersink) (Ra)

No burrs in metallic materials (hburr)
No delamination in composites (entry or exit) (Idelam)
No fiber fraying in composites (entry or exit)
No metallurgical change from excessive heat

Figure 2.9: Graphical view of the damage-related hole requirements

The drilling media or machine used must be able to repeatedly meet the defined specification
on each of the above criteria. Within the aerospace industry, there are two principal methods
available that are used for drilling assembly holes. Automated drilling using CNC machines
are increasingly being used for drilling holes in subassembly environments. Portable
equipment is extensively used in both the subassembly and final assembly stages (e.g. level
four assemblies in Section 2.1.2).

In most traditional industries, precision holes can be successfully drilled with a drill press
(pillar drill, milling machine etc) or CNC machines (e.g. level one assemblies in Section
2.1.2). However, since a significant number of aircraft structures are too large, complex and
irregularly shaped to be taken to a machining centre, portable precision drill motors (tools)
must be brought to the aircraft itself.

29
Aircraft Production

The wide range of hole sizes, the critically close tolerances required and the divergent
materials used in the aerospace industry demand that the portable drilling machines be
available with a broad range of cutter speeds, feed rate combinations and physical properties
that can accommodate virtually any workspace or application.

Temporary Fastening
Fixtures cannot fully ensure relative part location throughout the drilling sequence, whereas
parts to be drilled can gradually move out of position as the number of drilled holes increases.
This is caused by many factors. The most common factors are forces during drilling and
vibration of the machining. If fastening is done gradually along the drilling sequence, the
effect of parts gradually moving will increase. Temporary fasteners can hinder this
phenomenon. Temporary fasteners also prevent gaps between the parts in stacks to be drilled.
In general, a temporary fastener is inserted into every second or fourth hole. A temporary
fastener usually has a wedge lock mechanism, and can be inserted and tightened manually or
automatically. Figure 2.10 shows a fastener for manual installation:

Figure 2.10: A temporary fastener for manual installation

Temporary fasteners are also used if pilot holes are used. Pilot holes are used when details are
pre-assembled and holes are positioned. The pre-assembled parts can then be moved to an
automated or manual station for drilling the final accuracy holes. Insertion of temporary
fasteners is a time-consuming process. This is especially the case when drilling starter holes
up to a large size diameter or when reaming. In this case, the temporary fasteners have to be
gradually removed and inserted again.

Deburring
One of the worst enemies in the battle to maintain aircraft quality is fatigue cracks. Aircraft
fly for tens of thousands of hours at high altitudes where nothing is allowed to go wrong. An
old clich is that a car can pull over and call for a tow truck if the door falls off, but an aircraft
cannot, given its location ten thousand meters above the earth. Nor is a car exerted to the long
periods of vibrations such as those caused by the turbulence on an aircraft traveling at a sub-
sonic cruising speed. This is one of the reasons why aluminum in airframe assembly is not
welded with the same techniques used for car assembly. Instead, aircraft are mostly
assembled through drilling and fastening. It should also be mentioned that friction steer
welding actually is currently being performed in the assembly of aluminum parts for
aerospace applications. In the drilling process, if the parts in a stack are not clamped together
enough, there will be small chips/burrs pressed in between the parts. These chips may cause
cracks over time. To avoid this from happening, the parts are disassembled after drilling and
deburred. In manual drilling, deburring must always be done, but in automated systems the
clamping may be enough, and deburring can be avoided. Disassembling the parts for
deburring is a very time-consuming process.

30
 Aircraft Production

Sealant Application
As mentioned earlier in this chapter, some sections of the aircraft, such as the wings, contain
fuel and thus must be sealed. Other sections, such as the cockpit and cabin, must be air-tight.
In some cases, sealant must be located between parts in the stack to avoid them from rubbing
against each other and causing cracks over time. Sealant is either a liquid-based resin, or
something that can be taped on the parts. In many cases, the sealant is also applied on the
fasteners before insertion. In some cases, the sealant is applied before drilling, where one of
the parts in the stack has a thin film to avoid the sealant to stick on one surface to enable
disassembly. If the clamping force is not enough, and hence deburring is required, the parts
must disassembled. Drilling through sealant can cause problems, such as sticking to the drill,
or hampering the chips from evacuation from the hole. In other words, applying sealant is a
challenging operation for assembly automation.

Fastening
Fastening is the last step in the assembly process before removing the finished product from
the fixture. Fasteners used for fastening are rivets, screws, or bolts. There are many kinds of
fasteners; the most commonly used is the solid rivet fastener that is either hammered or
squeezed into place. Hi-Lok fasteners are used in stacks, especially where there are higher
demands on mechanical properties in the joint. Lock-Bolts and Hi-Loks are used similarly,
and both Hi-Loks and Lock-Bolts require access from both sides for swaging the collar, as
shown in Figure 2.11. Anchor nuts are a kind of screw fasteners that are used in the case of an
inside closed section, such as a wing box. Typically, Anchor nuts are used in interchangeable
joints, such as access panels etc. Blind rivets are the easiest fasteners from an automation
perspective, but for low strength requirements only. Today, however, blind rivets have been
developed that need only access from one direction, and yet still have the same joint strength
as Lok-Bolts and Hi-Loks fasteners.

Figure 2.11: A High-Lok fastener with the Collar being inserted.

Due to the stress-increasing characteristics of holes in aircraft manufacture, as well to the

large number of drilled holes on each aircraft, it is vital that quality holes with tight tolerances
are produced to reduce the chance of material fatigue and catastrophic failure during flight.
Therefore, the selection of rivets depends on Finite Element Analysis (FEA), which helps to
determine the requirements on stiffness and load that the product will be exposed to. From the
analysis, the fastener type is decided as well as if there is a need to have interference or
clearance fit.

2.5 Machines for Aircraft Automation

This Chapter will now moving further to present different machine concepts for automating
the airframe assembly process.

2.5.1 Large Scale Automation

The first solution in automation, for the riveting process, was based on machines built around
a C-frame structure (Fayad et al., 2002). In C-frame machines, clamping of pieces is
performed at the drilling location from both sides by bushings surrounding the drill bit. No

31
Aircraft Production

gap can develop during the drilling, and no disassembly and deburring operations are needed
prior to rivet/fastener installations (Sarh, 2002). A typical C-frame machine has one yoke arm
on each side of the airframe with one clamp table on each of the yoke arms to create the
clamp-up pressure as illustrated in Figure 2.12. The yoke arms can be manipulated up and
down and feed towards the airframe panel. The C-frame guarantees stiffness in the machine.
Machine-driven assembly can be carried out much faster, applying higher drilling and rivet
installation forces than in manual assembly with the limited capabilities of humans (Sahr,
2002). The C-frame drill/fastening machines have been used for many years for Level One
Assemblies presented in Section 2.1.2.
C-frame machine
One yoke arms on
Fixture with clamps each side of the
airframe structure

Travelling rail for

the C-frame
machine

Figure 2.12: A typical layout for a C-frame drill/fastening machine

A number of different auto-riveters are available, such as those manufactured by Gemcor,

Brotje and Electroimpact (Webb et al., 2003). One example of an auto-riveter machine is the
Electroimpact E4380 used for drilling and fastening both upper and lower surfaces of wing
panels on the A380 at the Broughton facility of Airbus UK. The machine has a range of 4m
vertically, 1.67m horizontally and 175 meter in length, and each weighs of 160 tons. Airbus
UK purchased four of these machines, and the first two were commissioned in April, 2003
(Zieve et al., 2004). The E4380 is sometimes a necessity in an industrial field such as
aerospace, due to the large work envelope and high demands on accuracy.

An example of a ring riveter machine, used to manage Level Four Assemblies in Section
2.1.2, is the SAAB ARAS (Automatic Riveting Assembly System), a fully-automated riveting
system developed by ATLAS COPCO Sweden for aircraft manufacturing, see Figure 2.13.
The ARAS is a vibration damped rivet-hammer substitute rivet squeezer, and it guarantees a
reduction of requiring riveting driving forces. The ARAS machine was used for the assembly
of the Saab 340 and Saab 2000 fuselages.

32
 Aircraft Production

Figure 2.13: The ARAS cabin system when it is positioned in the parking/turntable
location between the two jigs

A total of 13 different panels (top, side and bottom panels) were assembled in the system. The
ARAS system has no jig connected to the machine, where the panels have previously been
preassembled in jigs. This preassembly concerns locating the fuselage panels in the accurate
position using tack rivets. After this stage, the fuselage sections are brought in to the ARAS
machine, and the automation of drilling and fastening is engaged for the large number of
holes.

As Sarh (2002) stated, machines such as the C-frame have been developed for fuselage and
wing assemblies that are growing in size to exorbitant proportions in order to satisfy the
requirements of larger and larger structures. Sarh continues that high costs are dictated by
massive kinematics and complex controls that provide stability, precision and process speed.
The requirement for this stems primarily from the need to carry mechanical forces around the
part, from the upper to lower tool along the C-frame, gantry, yoke, bridge etc.

Crothers et al. (2002) state that aerospace has tended to utilize large, heavy and custom-made
equipment for the characteristics of accuracy, scale and process force. The expense of these
systems can only be justified on select, high-volume and large-scale operations.

Webb et al. (2003) claim that these kinds of fixed airframe automation machines are
expensive and inflexible, costing between 2.5 million and 3.0 million; they also require
custom-built foundations and require large areas of factory space. MacPhersson (2003) states
that the 5-axis machines for airframe assembly could cost upwards of US$ 3 million.

2.5.2 Parallel Kinematic Machines

Another more versatile robot concept, one combining more flexibility in the work envelope
but still keeping a stiff structure, are Parallel Kinematic Machines (PKM). Whinnem and
Nystrom (2000) claim that serial-linked architecture robots, as were presented in the next
section (Section 2.5.3), have problems with stiffness, and from that perspective parallel
kinematical machines has been tried as a solution. PKM has all links put in parallel between a
base and the TCP, which provides for better stiffness and ensures that the moving mass can be
kept small. The small moving mass itself enables the PKM to move at a higher speed and with
a low servo torque. The technology of PKM has been known for many years. The first fully
parallel structure including a 6DOF parallel kinematic platform was developed by Dr. Erik
Gough, who built a solution for testing tires for cars using a PKM (Gough and Whitehall,

33
Aircraft Production

1962). The solution from Dr. Gough was then applied as a flight simulator platform by D.
Stewart, and the design became known as the Stewart Platform (Stewart, 1965). The
challenge in making PKMs work more effectively is the extremely advanced controller
needed to manipulate the joints. The inverse kinematics, which are needed to calculate the
joint angles required to move its TCP to a defined point in the Cartesian space, does not have
a systematic calculation technique that is available as the Denavit Hartenberg used for serial
linked robots. Denavit Hartenberg has standardized calculation technique that restricts how
the kinematics for a serial linked robot is described. In a PKM, the inverse kinematics depend
on the type of machine elements that are used in each particular parallel machine.

An even bigger challenge is the fact that there is no solution to the forward kinematics; in
other words, moving one leg slightly will simultaneously affect all the other legs. The only
way to find a solution to the forward kinematics is to iterate using the inverse kinematics
equations. This is a CPU-consuming process, and has been one of the reasons that PKMs are
only now in recent years beginning to make the breakthrough. Fortunately, the most common
way to move the robot in a process is by using the inverse kinematics, such as moving the
TCP on a linear trajectory. Forward kinematics are only necessary to perform in some discrete
situations, such as emergency stops, search stops, start-up of the system etc.

One well-known PKM solution used in industry today, and especially in the aerospace
industry, is the Tricept Robot as shown in Figure 2.14.

Figure 2.14: A Tricept 605 robot drilling holes and installing anchor nuts at
Saab Aerostructures (Source: Saab Aerostructures)

The Tricept robot concept was patented by the Swedish company NEOS Robotics in 1988
(Neumann, 1988). The Tricept robot features a center pillar as well as three linear actuators,
plus three additional axes at the wrist. Tricept was first sold by NEOS robotics, and was for a
number of years manufactured by a company called SMT-Tricept in Vsters, Sweden; today,
however, Tricept is owned by the Spanish company Loxin. ABB sells a Tricept concept called
IRB 940 using a standard S4Cplus robot controller.

34
 Aircraft Production

The accuracy of a Tricept robot is 10 times more precise than a traditional industrial robot,
with a positioning accuracy of +/- 0.2 mm and a repeatability accuracy of +/- 0.02 mm (ABB
IRB 9400, 2002). There also exists a kind of standardized metrology-integrated solution for
Tricept robots and two digital cameras positioned between the workpiece and end-effector
called TI2. The TI2 solution is presented in Section 5.3. The work envelope for the IRB 9400
is shown in Figure 2.15.

A Absolute tolerance area

B - Work envelope

Figure 2.15: The working range (ABB IRB 9400, 2002)

Note in Figure 2.15 that the work envelope is in the range of 1600 mm in width and 600 mm
in depth, where the work envelope with the guarantied positional accuracy of +/- 0.2 mm is
only a window of 700 x 350 mm.

There exist several examples where Tricept robots have been used in aerospace applications,
both in industrial installations and in research labs. One major aerospace installation of the
Tricept robot concept is seen at Airbus Spain at the Getafe site. An example where extensive
research is currently being carried out on Tricept robots for installing solid rivets is at
Nottingham University. There, the application of the Tricept robot concerns the handling,
drilling and squeezing of solid rivets for aerospace applications together with Bombardier
Aerospace Short Brothers Ltd, Comau Estil and ATA Engineering Processes Ltd (Webb et al.,
2003).

2.5.3 Serial Kinematic Industrial Robots

A serial kinematic Industrial Robot is based on having a
base and links connected in a serial chain to the TCP.
Joints can be revolute, where the links are connected by
a rotational joint, or prismatic, where the links are
connected by a linear joint. Industrial robots often have
the joint variables and link parameters defined
according to the Denavit Hartenberg notation (Craig,
1989). The movement of the TCP in Cartesian space is
calculated to joint values according to the inverse
kinematics. Encoders often are installed directly on the
servos. The servos are connected to the joints through a
gear box.
Figure 2.16: An
articulated robot with six
35 revolute joints
Aircraft Production

One typical robot architecture is the articulated design, shown in Figure 2.16, is one that
usually consists of two shoulder joints (one for rotation about the vertical axis and one for
elevation out of the horizontal plane), an elbow joint whose axis is usually parallel to the
shoulder elevation joint, and two or three wrist joints at the end of the manipulator (Craig,
1989). Articulated robots sometimes go under the name jointed, elbow or anthropomorphic
manipulators.

Since the introduction of teach-and-repeat robots in the 1960s (the first commercially-
produced industrial robot was known as the Unimation), robots have been used in
manufacturing (Bolmsj, 1989), the most common areas being spot welding, arc-welding,
spray painting, deburring and pick-and-place automation. The largest application field has for
many years been the automotive industry. During the period 1990 2004, the number of
robots manufactured has ranged from 60.000 to 99.0000 per annum. This included a peak at
year 2000, a large drop between 2001 and 2002, and a strong recovery in 2003 with a total of
81.800 units sold (World Robotics, 2004). Within a historical context, there are now about
900.000 robots (mostly Cartesian, gantry robots) in use around the world (World Robotics,
2004). Industrial robots have been used to a great extent in the Automotive Industry.

Normally, robots come with a special controller that

interprets the programs (see Figure 2.17). The
controller calculates robot movements that are
executed by electro-mechanical motors. Encoder data
and I/O is used to communicate back and forth with
the robot. These controllers are highly advanced
industrial computers that require special design with
security, EMC and stability issues in mind. The
controller can be connected to other systems either via
Figure 2.17: An Industrial
Ethernet (TCP/IP) or serial communication (RS-232).
Robot and Controller
The robot comes with a software package called RobotWare, which is divided into three
groups:

BaseWare OS
BaseWare Options
ProcessWare

BaseWare OS is the robots operating system, and constitutes the kernel of the RobotWare
family. BaseWare OS provides all the necessary features for fundamental robot programming
and operation, and is an inherent part of the robot but can be provided separately for
upgrading purposes. BaseWare Options are options that run on top of the BseWare OS and
provide for additional functionality, such as multi-tasking, transferring of information from
file to robot, communication with a PC, performing advanced motion tasks etc. ProcessWare
also runs on top of the BaseWare OS, and is designed for specific process applications like
welding, or painting. It is primarily designed to improve the process result and to simplify
installation and programming of applications.

The BaseWare Options installed in the robot used in the research presented in this dissertation
were:

36
 Aircraft Production

Advanced Motion
Multitasking
FactoryWare Interface
RAP Communication
Ethernet Services

Some of the Option packs installed in the robot are needed for the integration presented and
further explained in Chapter 6.

The work envelope of the ABB IRB 4400, 60kg payload robot which was used for this
research is presented in Figure 2.18.

Figure 2.18: The working range for the ABB IRB 4400/60
(ABB IRB 4400, 2004)

The following section explains how these kinds of robots are suited for the airframe assembly
application.

2.6 Industrial Robots for Aircraft Automation

The aerospace industry has now begun to realize the potential of using industrial robots for
automation. The automotive industry has also placed new demands on robots to handle
increasingly complex tasks requiring greater precision. Among others, these two scenarios
have led to robot suppliers increasing the capability of their robots. This section will discuss
the current capability of industrial robots versus the new requirements for their use in the area
of aircraft automation.

2.6.1 Accuracy
An industrial robot is developed to conduct repetitive tasks such as spot-welding or pick-and-
place operations. For that purpose, robot manufacturers have strived to improve repetitive
accuracy, which is a measure of a robots ability to return to a previous position. Today, it is
still common to use a joystick to program the robot a so-called handheld tech-pendant. The
tech-pendant is used to manipulate the robot and store locations, hence the name teach-in
programming. In teach-in programming, all the errors in the robot are eliminated by the eyes
of the operator (Whinnem and Nystrom, 2000). Higher utility of CAD simulation systems has
introduced larger utilization of offline programming systems for programming a robots paths.

37
Aircraft Production

(Offline programming is presented in Chapter 8.) When programming the robot offline, the
error correction by eye principle cannot be used. In this case, the robot is given a frame with
position data for x, y, and z as well as the additional angles. In this situation, all the errors in
the robot will affect the final position of the TCP. This is called absolute accuracy, or
sometimes positional accuracy. Absolute accuracy has been defined, according to (Helin,
2002), as follows:

A measure of the difference between the actually attained position of the robot
end-effector and the commanded robot controller expressed in the current object
frame.

In the manuals for ABB robots, absolute accuracy is not guarantied to be better than 8-15 mm
(ABB acc, 2004). This has been confirmed by experience at VOLVO Cars, which postulates
that a standard, large-sized robot has positional accuracy of around +/- 5 to 10 mm Axelsson
(2002). The reasons for errors in the robot include:

Geometric errors such as error in link lengths, constant offset angles, angular offsets
on actuated joints
Non-geometric errors such as structural joint/link compliance
Gearbox associated backlash and inaccuracies

The accuracy can be improved by either updating the robot parameters in the controller or
offline filtering of TCP (or joint) trajectory data using pre-identified error models. Monsarrat
et al. (2004) state that these calibration packages typically allow an improvement in robot
accuracy in the order of 0.8-1.0 mm.

This accuracy meets the requirements for typical robotic applications such as spot-welding
and pick-and place operations for the automotive industry. Positional accuracy for spot weld
in car assembly are allowed do deviate +/- 2.5 mm (Axelsson, 2002).

In the aerospace industry, however, the requirement for accuracy is in the order of 0.05-0.5
mm, depending on the application. In aircraft assembly, parts require positional accuracy in
the order of 0.2 mm. In drilling, hole accuracy may require positional accuracy of 0.05 mm
if the parts being drill are interchangeable, which means a substructure being drilled must fit
any mating part. The opposite is called stack drilling, where the two parts to be assembled are
drilled together in the same drilling cycle, allow much lower demands on accuracy.
Essentially, this means that if a robot is purposed to drill interchangeable parts, the robot need
to undergo extensive calibration. If demands on the accuracy of drilled holes were lower,
todays robots could be used. Successful applications of robots being used for aircraft
assembly today are presented in Russel and da Costa (1999) and DeVlieg et al. (2002).

Offline filtering is a method to make a static calibration of an industrial robot using a

metrology system to update the calibration file in the robot controller. Monsarrat et al. (2004),
from the National Research Council Canada, showed a successful attempt to increase
accuracy to the level required for aerospace applications. The NRCC team managed to get the
absolute accuracy of an ABB IRB 4400 (as used in this research), from a maximum Cartesian
error of 8.2 mm to 0.4 mm using their own method and a Leica LTD500 tracker.

Another successful aerospace example of offline path correction was developed by the
Australian ADTACC project, called the Automated Path Calibration System (APaCS), an
adaptation of Iterative Learning Control. The method adjusts the virtual world to the real
38
 Aircraft Production

world (Crothers et al., 2004). This approach relies on extensive measurement of robot
trajectories where robot targets were moved as a result from the robot deviating from the
nominal path. The application was trimming of detail components where the method gave an
accuracy which bettered the target of 0.5 mm. The solution was brought in to production
2004.

In the ADTACC project, the University of Wollongong was responsible for the development
of a micro-positioning end-effector. Essentially, this end-effector gave the robot three extra
degrees of freedom of fine resolution. By using two of the electric actuators (the third being
for drill feed) the drill TCP could be guided by a metrology system to a desired point within
an error as low as 0.007 mm from the metrology system measurement (Crothers, 2005). This
system is to be brought in to production in 2005.

Helin et al. (2002) have presented tests results from ABBs High-accuracy program which
clearly show that it is easier to reach higher accuracy in smaller robots. ABB used their
CalibWare application to collect measurement data using a LTD500 robot, and through a
standard least square optimization identified the set of parameters that minimized the
deviation between the commanded and attained end-effector positions. ABB concluded the
error to deviate as +/- 0.2 mm for the smaller robots and +/- 0.6 mm for the larger robots
(Helin et al., 2002). This option pack can be purchased for ABB robots. The Option pack is
called Absolute Accuracy (ABB-ABS acc, 2004), which is an accessory that can be purchased
along with a new robot from ABB.

2.6.2 Dynamic Loads

The previous subsection presented some examples of extensive calibration procedures used to
increase the default absolute accuracy in industrial robots. The author calls these calibration
methods static, which essentially means the improvement to robot accuracy will be valid for a
static scenario. This section will discuss this matter, and especially if the scenario is not static.

The ABB IRB 4400 used in this research can include a parameter for the amount of load
attached to the TCP, called GripLoad. The load is in the gravitational direction; hence, if the
force vector is from a direction other than straight down to the floor, the robot controller can
not cope with that force. The accuracy of a robot is load-dependent, which means that the load
data must be correctly defined in the controller (Axelsson, 2002). Axelsson (2002) further
explains one of the problems with robots in spot-welding. Axelsson explains that the arms of
the welding gun are exposed to a number of forces during the welding process, where the
electrode is tip-dressed at certain intervals. VOLVO Cars has solved the issue by having the
guns free-floating (equalizing) in one direction at the moment of welding due to the
geometrical deviation in the robot caused by the dynamic forces.

In drilling using an industrial robot, forces change rapidly. Today, suppliers for drilling end-
effectors handle this problem either by pre-pressurizing the drill bushing that the cutter is fed
through, or using an additional pressure foot on the drilling end-effector. This introduces
forces to the robot. To eliminate disturbances on the robot, the pre-pressure force must be
significantly higher than the thrust force in the drilling, which ensures that the drilling end-
effector maintains its position and orientation during drilling.

In drilling and fastening applications, the actuating force is in the normal direction from the
pre-pressurized surface. Dgoulange et al. (1994) stated that when a robot is exposed to a
force at the TCP, it will not stay orthogonal to the work piece anymore. Experience in their
research has shown that the robot will change the TCP in 6 degrees of freedom as the one-
directional force exerts on the TCP.
39
Aircraft Production

When installing a drilling end-effector on the TCP of the robot, there are two principle
different installation alternatives to choose from. Either the end-effector is installed with the
cutter pointing straight out in the same direction as axis 6 as the robot (Figure 2.19a), or
rotated 90 degrees to the first alternative (Figure 2.19b).

Figure 2.19a: The Figure 2.19b: The

pointing configuration hanging configuration
Figure 2.19a above shows the pointing configuration, while Figure 2.19b the hanging
configuration. Traditionally, the end-effectors are installed with the hanging alternative. In
choosing that alternative, several issues should be considered. With the hanging
configuration, joint 5 of the robot is rotated 90 degrees, and as the pre-pressure is applied
prior to drilling a high torque is placed on that joint,
which will cause the joint to deflect and loose the
perpendicular direction to the surface (loss of
normality to the hole direction). In the pointing
configuration, the robot can push with a stiffer
configuration. In addition, if the weight of the end-
effector is within the robot payload limits, the robot
will manage to carry the load using commands such
as GripLoad for ABB robots. When choosing the
pointing configuration, there is one important factor
to pay attention to the center of gravity for the end-
effector which is not linear to the distance out from
TCP-zero. Figure 2.20 shows the relationship
between the allowed distance of center-of-gravity of
the payload and the distance away from TCP-zero for
an ABB IRB 4400 60 kg payload robot. Using the
IRB 4400 as an example, the distance relationship to Figure 2.20: The load diagram
payload must be considered in the design of the end- for an IRB 4400/60
effector and the selection of configuration.

Another issue that is becoming more evident in using the pointing configuration is the fact
that when joint 5 is in zero degrees, it causes axes 4 and 6 of the robot to point on the same
line; the result is that the robot ends up in a singularity. A singularity in this case can be
explained in this way: there are infinite numbers of solutions in the inverse kinematics to
calculate joint angles from a given TCP location; this will make the robot go in to an error
mode and stop. This situation, however, can be avoided using offline programming systems,
where the singularities are detected offline, hence avoiding unnecessary production stops.
Moreover, choosing the pointing configuration for attaching the end-effector to the robot is
the same as eliminating joint 6. The manipulation of joint 6 will not alter the location of the
drilling machine relative to the hole, since the cutter itself is rotating. This will limit the
flexibility of the robot for reaching a large work envelope as well as the number of
configurations the robot can manage using joint 6 to manipulate the end-effector.
40
 Aircraft Production

In a conclusion to the selection of pointing- or hanging configuration of the end-effector, it

stands clear that the pointing configuration gives the robot a stiffer mode, but limits the work
volume. Choosing the hanging alternative gives the robot a larger work envelope, but
consideration must be given to stiffness problems.

In a drilling application, there are two domains where the forces change the first domain
before drilling when pre-pressurising the drilling machine to the workpiece surface, and the
second domain varying forces during drilling. The pre-force is static, and the forces during
drilling are dynamic. Using a metrology system, the static deviation can be identified and
adjusted for, but the dynamic forces are more challenging. Using a metrology system in
conjunction with a robot to compensate for dynamic forces, the metrology system would need
to interact with high speed with the robot. This problem is further explained in Chapter 5
(metrology). Chapter 3 also presents a drilling method that has low thrust forces during
drilling in order to avoid this problem.

In addition to drilling, Chapter 4 presents research using the IRB4000 to reconfigure a fixture,
which is yet another example of static forces. Chapter 2 presents how this situation is solved
using a metrology system to ensure high positional accuracy although unknown forces.

2.6.3 Work Envelop

The work envelope of an airframe structure or substructure is in most cases larger than the
working range for a robot. This has been yet another reason for using large machines. It is
possible to put the robot on a robot travel track. The problem using a travel track is that that
movement is only along one horizontal direction. Moreover, using travel tracks for robots,
there are two important factors to consider. The first one is that the positional accuracy error
of the track adds to the accuracy error of the robot. For example, a traveling track from ABB
the ABB RTT 1400/2400L has a positioning accuracy of +/- 0.05 (ABB RTT 1400/2400L,
2002). In addition, the robot is not fixed to the floor using a travel track, which may lead to
problems with stability.

2.6.4 End-effectors
In robotics, an end-effector is a device or tool connected to the end of a robot arm. The
structure of an end effector, and the nature of the programming and hardware that drives it,
depends on the intended task.

Aerospace assembly is primarily focused upon drilling, fastening and the accurate movement
and locations of parts. A way to handle the low accuracy of industrial robots is to design end-
effectors with multi-functionality. Drilling a hole and then change to a fastening end-effector
and go back to the hole is difficult. The requirements on repositioning the end-effector are in
many cases to demanding for the robot to manage. One alternative is to keep the robot in one
location and perform all the assembly operations, where the end-effector comprises all
functions necessary (e.g. drilling, probing, applying sealant, fastener insertion and fastening).
Two common problems in using multi-end-effectors are large weight and a high expense. The
other alternative, however, is hampered by the issue of the robot returning to the same
position using several smaller end-effectors.

A current trend in the aerospace industry is to buy large robots for handling multi-end
effectors weighting several hundred kilograms. Using large robots for aircraft assembly
conflicts with the fact that larger robots have inferior accuracy compared to smaller robot.
This matter was discussed in Section 2.6.1. This indicates that end-effector suppliers should

41
Aircraft Production

consider building smaller end-effectors if they are to be fine-adjusted, which would require
smaller robots. A question being posed to the end-effector suppliers today is that of how large
end-effectors can be positioned with higher accuracy.

2.6.5 Final Remarks Using Robots for Airframe Assembly

This chapter has shown that robots have a potential to match the requirements for airframe
assembly. It is not realistic that industrial robots should completely replace large scale
automation machines; whereas the C-frame machine would require one robot on each side of
the structure working in unison to form the C-frame-like structure that is needed to press the
solid rivets, Lok-bolts etc. There is no doubt that industrial robots can do that, but not with the
fast cycle times seen in C-frame machines.

Offline filtering techniques improves accuracy of robots for a static load. Drilling, however,
provokes load to the robots, which is not gravitation load, but process load that can have any
direction; hence drilling is a challenging field for industrial robots. In the next chapter, a new
drilling technology is presented that has low cutting forces, and is therefore called an enabling
technology for robotized drilling. If forces are varying in the application, however, and
especially if that variance cannot be modeled, then the robots either have to measure that
force to cope for it or measure the deviation from the nominal position. Chapter 5 presents a
technique to measure how the robot deviates from the nominal location using a metrology
system.

Figure 2.21 illustrates the different levels of automation for airframe assembly. Large-scale
automation machines, which drill and fasten at high speed, are placed to the right of the
assembly concept diagram. These machines are expensive. In addition the fixture is many
times part of the machine, which further increases the complexity and thus the cost. These
machines are described in further detail in Section 2.5.1. Applications with a low number of
fasteners per year are typically done manually by an operator using handheld machines for
the drilling and fastening operations. It is the cheapest solution in an investing volume
context; hence, it is located down and to the left in Figure 2.21. The author puts industrial
robots in the middle between the heavy automation solution and handheld solution.

Cost
Heavy
Automation

Industrial Robots

Handheld

500.000 1.000.000
Fastener Count / Year
Figure 2.21: Assembly concepts in relation to investment
and production volumes

The task of the author has not been to map these different areas, but instead to try and expand
the area of applicability for industrial robot use. This dissertation presents ideas on how
robots more quickly and precisely can be started up, and how robots can be used for
automation although small product volumes. These parameters expand the dashed ring down
to the left in Figure 2.21. Moreover, if robots could work quicker and more robustly, the

42
 Aircraft Production

dashed ring could also be expanded down to the right without increasing the investment costs,
but yet increase the assembly operations frequency.

The author postulates the reasons industrial robots have not yet made their breakthrough in
aerospace manufacturing as the following summary:

Poor robot accuracy

High accuracy non-contact metrology was not existent
Poor provision for sensor feedback and system integration
Limited 3D-CAD and offline robot programming software
Low product throughput
Limited advances in flexible jigging
Low speed, high force machining
Conservative aerospace culture

In this research the developing and building of physical demonstrators has been considered an
important method to prove that the new technology will work in a real production
environment. Physical demonstrators are also an important way to marketing new
technologies to airframe builders. New sensor technology being incorporated in the robots to
improve their capability is one way to increase accuracy and achieve higher robustness and
capacity. Case studies, using state-of-the-art offline programming software for operation
planning, will help to realize the potential for complete system programming. New, low cost
flexible tooling systems that can handle changeovers with fewer systems, and thus keep
investment volume low and total throughput high, are needed. New drilling technology can be
a way to make it easier to use robots for drilling automation.

Although industrial robots has been brought forth for automating high-volume production,
aerospace applications with their characteristic low-volume product scenarios still can benefit
from this technology. Probably, in comparison with the high volume scenario having robots
bolted to the floor making a few operations in an aerospace low-volume scenario, robots are
likely to be mobile to move around in the factory to increase their utility factor. Alternatively,
product units of different types are moved into a robot cell. The fact remains: industrial robots
for aerospace must be more flexible in its utilization than for a high-volume scenario.

An industrial robot is a rather standardized piece of production equipment and can easily be
programmed in an offline-programming (OLP) systems, and hence achieve low lead times.
OLP also enables programmers to use non-manufacturing dedicated programming languages.
In fact, OLP systems automatically process the dynamic simulation into specific robot
language; hence, those working in process planning can easily integrate robots from different
suppliers into one simulation program without knowledge of the robots dedicated program
language. (This is more thoroughly presented in Chapter 8.) World Robotics (2004) has
summarized a number of surveys conducted among robot users and prospective users, and has
come up with the following motivations for investing in industrial robots (not necessarily in
order of ranking):

Reduce labor costs

Improve product quality/obtain a more even quality
Improve quality of work by eliminating hazardous, heavy and/or repetitive work
cycles
Increase output rate
Increase product flexibility

43
Aircraft Production

Reduce material waste

Comply with safety rules
Reduce labor turnover/difficult to recruit workers
Reduce capital costs (inventory, work-in-progress, floor space etc.)

Dgoulange et al. (1994) state that efforts to use standard industrial robots (IRBs) have
increased and are currently being tested, due in part to their high flexibility. Using
standardized production equipment, such as industrial robots, is many times less expensive
than using large NC-machines, which are used extensively in todays airframe automation.
Industrial robots are not dedicated for one specific purpose, whereas they can be used for
drilling, the same robot can also be used for other processes, such as pick-and-place operation
or rivet insertions and fastening. Due to its versatility the robot is not exclusive for one
product, but can be applied to many parallel production flows. Calibration for numerical
machines are traditionally calibrated in its full work envelope, whereas an industrial robot can
be calibrated for a smaller work window where requirements are high, and this work window
can be moved by simply changing calibration parameters.

What will be shown further in this dissertation is enabling technology for using industrial
robots for airframe assembly. These enabling technologies do not rule out being applicable for
PKM solutions, such as the Tricept robot. Many times in research you take what is at hand.
An ABB IRB 4400 was available for this research and was taken on. The author will not rule
out Tricept robots as a solution for what is presented in this dissertation. The enabling
technologies could in fact be directly applied to a robot type like Tricept or any other robot
type available without extensive work. The most evident downsides using a Tricept robot
instead of an industrial robot is work envelope, access, and cost. The work envelope with
assured accuracy was very small, which would require sensoric interface anyway. An
anthropomorphic robot arm can reach more complex structures, not being hampered with
parallel arms. Tricept robot are not as mass-produced, hence, more expensive compared to a
conventional industrial robot.

44
PART II: AIRFRAME ASSEMBLY FOR AFFORDABLE
AUTOMATION

PART II
AIRFRAME ASSEMBLY
FOR AFFORDABLE AUTOMATION

In Part II, the frame of reference for airframe assembly is presented, which will
give the reader an understanding of the general aspects of this field. Part II
introduces typical machinery for airframe assembly. Finally, the challenges and
requirements for using industrial robots for airframe assembly are discussed to
support the authors approach in developing the key enabling technologies that
are later presented in this dissertation.

45
46
 Robotic Orbital Drilling

3 Robotic Orbital Drilling

Following the introduction on aircraft production, Chapter 3 continues with the presentation
of a new drilling technology, evaluated as part of this research as one approach to
simplifying the use of industrial robots for drilling in the aerospace industry. In Chapter 3,
conventional drilling is used as a reference drilling method for orbital drilling, presented
later in this chapter. In Chapter 2, an introduction to typical airframe assembly was
presented, where drilling was one of the process steps. This chapter will further present the
drilling process in order to provide the reader with a deeper understanding of the drilling
process as well as the conclusions drawn on orbital drilling by the author at the end of this
chapter.

3.1 Conventional Drilling

The concept of drilling constitutes all methods associated with the machining of cylindrical
holes. Conventional drilling, which is the most common method for drilling, is characterized
by having a stationary tool center. The main movement is the rotating movement performed
by the tool (sometimes called cutter). In this dissertation, feed rate is defined as the straight
axial movement of the cutter. Feed rate is generally caused by the operator, or by the drilling
machine in more advanced applications. The cutting speed is the combined peripheral speed
and axial feed rate towards the workpiece. Both peripheral speed and feed rate can be varied
to suite cutting characteristics when machining different materials.

Figure 3.1: A typical Conventional spiral cutter

A spiral cutter is the most common tool used in drilling (see Figure 3.1). The spiral cutter
constitutes one cutting part and one interface part for securing the tool in the drilling machine.
Some spiral cutters have standardized dimensions that can be purchased off-the-shelf. Spiral
cutters normally have a conical cutting edge and a slightly decreased diameter from the
cutting edge to the attachment section in order to reduce friction between the cutter and the
hole. Bushings are normally employed when using spiral cutters.

In manual drilling, it is common to drill one hole starting with a small diameter and then
increase the holes diameter gradually. This is done to ensure a high degree of final accuracy.
Gradually increasing the hole diameter is done by first using drilling templates to drill the
pilot holes. As the pilot hole has been drilled, the diameter is increased to its final diameter.
One example is shown in Table 3.1.

47
Robotic Orbital Drilling

Table 3.1: A summary of drill and ream operations at Saab Aerostructures (Bylund,
2001)
YAA 4/YAD 4 Method 1 2,6 mm 3,8 mm 4N10
YAA 4/YAD 4 Method 2 2,5 mm 3,8 mm 4N10
YAA 4/YAD 4 Method 3 2,5 mm 3,3 mm 4N10
YAA 5/YAD 5 Method 1 2,6 mm 4,8 mm 5N10
YAA 5/YAD 5 Method 2 2,6 mm 4,1 mm 4,8 mm 5N10
YAA 5/YAD 5 Method 3 2,5 mm 4,1 mm 4,8 mm 5N10
YAA 5/YAD 5 Method 4 2,5 mm 4,1 mm 5N10
YAA 6/YAD 6 Method 1 2,6 mm 5,7 mm 6N10
YAA 6/YAD 6 Method 2 2,6 mm 4,1 mm 5,7 mm 6N10
YAA 6/YAD 6 Method 3 2,5 mm 4,1 mm 5,7 mm 6N10
YAA 6/YAD 6 Method 4 2,5 mm 4,1 mm 4,8 mm 6N10
Drilling
operation
Reaming
operation

Note in Table 3.1 that each hole required drilling to be performed between two and three
times with increasing diameters. The last column for each hole in Table 3.1 corresponds to the
reaming operation.

3.1.1 Reaming
In an aircraft application, the demand on surface quality is higher than that which a
conventional spiral cutter can provide. In this case, a reaming tool is necessary to improve the
hole dimension accuracy and surface finish after drilling. The most demanding quality figure
in serial production is IT 7. The reamer tool constitutes one cutting part, one cylindrical part
and one attachment section. The cutting edge can be straight or have a spiral design. Figure
3.2 shows a spiral reaming tool. There are two variants of reamers: the hand held reamer and
the one suited for machines. The conical section of the tool guides the rest of the tool through
the material; it also helps avoid vibrations in the tool.

Figure 3.2: A reaming tool

There exist tools that contain both a drilling section and a reaming section in the same tool.
This enables drilling to be performed in one shot. The problem that can occur in this case is
that the guidance of the tool may be offset. The reason for this is that these kinds of tools are
more suited for stationary machines that are attached on a drilling jig.

3.1.2 Plug Gauges

The most common quality checks that are performed on Hi-Lok are tolerance, surface quality,
and hole straightness. Using plug gauges is a quick method to measure hole tolerance
according to the go/no-go concept. Figure 3.3 shows a typical plug gauge. Hole straightness is
measured using the plug gauge, and a hole angle measurement device attached to the plug
gauge.

48
 Robotic Orbital Drilling

Figure 3.3: A plug guage

3.1.3 Countersinking
To provide for a smooth aircraft surface some fasteners must be countersunk. This process is
done in a separate operation. For example, a Hi-Lok requires a countersink angle (see
Figure 3.4a) of 100. The countersink inspection is performed using static scales (see Figure
3.4b), with max and min intervals.

Figure 3.4a: Countersinking tool Figure 3.4b: Static scales

3.1.4 Deburring
When drilling a material stack, a pressure is always applied on the workpiece to remove the
gap between the parts prior to drilling. This is called the clamping force. If the clamping force
is not large enough, the parts in the stack will be pushed apart to a small extent and burrs may
be formed between the layers of the stack. These burrs must be removed before the parts can
be fastened, as burrs have been proven to initiate fatigue cracks. This operation requires the
disassembly of the structure, and the subsequent deburring of the workpiece using a deburring
tool.

If the clamp-up force is high enough, deburring is not necessary. The force required, however,
is far higher than a human can manage; this is why C-frame machines are used on a large
scale in aircraft production. C-frame machines, however, are expensive and require easy
access. Therefore, deburring is still a common operation today.

More information about conventional drilling in aerospace applications using spiral cutters is
summarized by Nichol (2001). Additional information concerning drilling in aerospace
applications using one-shot drill is presented by Fernandes (2005).

3.2 Orbital Drilling

From what first was known by the KTH-method (Person et al., 1997) was later named the
orbital drilling method, or just orbital drilling. The concept of orbital drilling is based on
using high-speed machining to eliminate some of the existing problems in conventional
drilling techniques.

Orbital drilling may appear very similar to circular interpolation, but the principal difference
is that the orbital motion is a mechanically forced motion, executed in a polar coordinate
system (see Figure 3.5). Thereby the orbital motion can be carried out at higher speed and

49
Robotic Orbital Drilling

with higher precision than what is possible using normal interpolation. The principle of
executing the orbital motion in a polar coordinate system also makes it possible to build small
drilling units for this method that are suitable to use in aircraft assembly operations, where it
is necessary to bring the equipment to the work piece rather than bringing the part to the
machine tool (Eriksson, 2005).

The first evident advantage over conventional drilling is the fact that the tool does not have a
stationary tool center. This chapter will present results from measuring forces in orbital
drilling to show that the thrust force is minimized and many of the problems associated with
conventional drilling (i.e a stationary tool center) are eliminated.

Figure 3.5: The Orbital drilling cutter orbiting the hole center

Moreover, with orbital drilling the tool diameter is less than the hole diameter, enabling
efficient chip and heat extraction. The cutting edge is partially and intermittently in contact
with the workpiece, allowing for efficient chip extraction. Chip extraction is also simplified
from the high spindle speed. In carbon fiber, the chips are in fact dust particles that are sucked
out by vacuum.

There exist two fundamental principles in orbital drilling. The first one is the Portable Orbital
Drilling Unit (PODU) unit, which is for use in manual drilling where the machine is attached
to drilling templates. One case using the PODU unit for SAAB Ericsson Space was presented
in Appended Paper VIII (Lindqvist and Kihlman 2004); this was a further development from
Lindqvist et al. (2001) and Appended Paper IV (Kihlman et al., 2002). SAAB Ericsson Space
implemented the use of two PODU units as one of the first customers to use the technology.
The second principle was the Computer Numerically Controlled Orbital Drilling Unit (CNC
ODU). In Paper VIII, the implementation and integration of the CNC ODU on a KUKA 200
robot was presented. These results are discussed in Section 3.4.2 of this chapter.

The difference between the PODU and the CNC ODU is the use of electrical motors instead
of pneumatics in the CNC ODU, whereas the settings of the machine can be adjusted
throughout the drilling cycle. The eccentricity is adjustable; thus the precision of the hole is
independent of the precision of tool. This in turn makes it possible to drill high precision
holes with low precision tools and to drill holes of different diameters with tools of one
diameter. As seen in Figure 3.6, orbital drilling can be used to machine different kinds of
axis-symmetrical holes such as cylindrical, conical and complex shaped.

50
 Robotic Orbital Drilling

Figure 3.6: Complex hole geometries in orbital drilling

The Orbital drilling cutting tool has axial and radial cutting edges. One edge is for performing
the milling operation, which cuts in the radial direction. The second one is on the face of the
tool, which cuts in the axial direction. A normal milling operation involves moving the cutting
tool in a radial direction, and drilling is obviously moving only in an axial direction. Orbital
drilling moves in a helical movement through the material.

3.2.1 Fatigue analysis

At Linkping University, a study was undertaken to assess the effects of hole quality on basic
material properties and to evaluate possible methods to make connections between properties
and surface quality (Johansson et al., 2002). To achieve this goal test, specimen were drilled
with two different drilling methods, traditional and orbital drilling, using different drilling
parameters. The drilled specimens have been tested under static tensile and compressive
loading as well as fatigue. The damage introduced during testing was studied by acoustic
emission. The hole quality of the drilled holes was also evaluated.

Conclusions on stress strain behavior revealed that there where certain differences between
the conventional and orbital drilling methods. Fracture stress strength, the first stepwise
deviation from the linear stress strain behavior, showed a slightly higher strength for orbital
drilling. The results from the fatigue testing showed a clear trend for plain holes towards
higher fatigue life for orbital drilling. This was partly explained from the higher surface
quality of the holes observed for orbital drilling. Although there were promising results for
straight holes for orbital drilling, countersunk holes gave disappointing results for orbital
drilling. This was partly concluded from the fact that the cutter was moved in a helical
movement through the material, whereas the cutting edge caused higher surface roughness for
the countersunk region in orbital drilling. It should be mentioned that Novator, the patent
holder for orbital drilling, later introduced a cutter with a countersunk angle on the cutter
providing high surface quality, with a slight loss of flexibility from using a static tool setting.
In addition to the strength tests comparison between conventional and orbital drilling, hole
roundness measurements were also undertaken, using a TALYROND roundness tester.
Surprisingly, the results showed better roundness in the orbital drilling method. As mentioned
earlier in this section, the surface quality in orbital drilling was concluded to be better than
that for conventionally drilled holes. These tests were undertaken using the Messma-kelch
robot. Orbital drilling has small striated concentric machining marks along the hole surface.
These pits or intrusions, caused by tearing fibers, are however only in order of 0.1-0.2 mm,
and that phenomenon is more frequent for conventionally-drilled holes, likely caused from
fibers and bundles of fibers torn out during machining (Johansson et al., 2002).

3.2.2 Cost modeling

In the ADFAST project, a cost model was implemented to evaluate various economical
aspects when buying new machinery for assembly or when changing existing applications.
With this model, the user can evaluate changes in cost with different volumes or different
equipment for a specific application.

51
Robotic Orbital Drilling

The cost model also enables the review of times and costs, providing individual contributions
to the overall costs. This research proved portable orbital drilling to reduce cycle time by half,
compared to conventional manual drilling methods. This conclusion rests upon the fact that it
is possible to obtain the required final hole in one shot. In addition, depending on the
application, orbital drilling can provide cost savings up to 50% when compared to
conventional drilling. These cost savings arrise especially from machining, consumables, and
temporary fastening costs. In this context however, it must be pointed out that orbital drilling
was still under development during this research, and therefore it is difficult to quantify and
compare fixed costs such as purchase of equipment. What is evident is the fact that the fewer
number of steps within the orbital drilling sequence have the greatest impact on the cost.

From an introduction to orbital drilling, this dissertation will now continue to address the
actual experiments have been undertaken in this research and the conclusions that have been
reached by its author.

3.3 Force Measurements

At the time this dissertation was published, orbital drilling was considered a new drilling
method. Research made by the author was to compare forces in orbital drilling with the
reference method of conventional drilling, and to conclude how orbital drilling renders
possible the use of robotized drilling. The conceptual idea was to start out measuring forces in
both methods conducted by the author. In Chapter 2, it was concluded that forces using
industrial robots are one reason why it is difficult to automate drilling. After the force
comparison between conventional and orbital drilling was undertaken by the author in a test
bed, the author then had the opportunity to mount the Portable Orbital Drilling Unit (PODU),
for the first time, on an industrial robot for evaluation. Forces were again measured, but in
addition the force sensor was used to investigate pre-loading etc. Furthermore, the deflection
was measured in the PODU when it was attached to the robot. The conclusion of these results
were published in Appended Paper IV, (Kihlman et al., 2002). It must be mentioned that hole
quality analysis was not performed at this stage. Hole quality in robotic orbital drilling was
later investigated in the integration between a CNC Orbital Drilling unit (CNC-ODU) using a
KUKA 200, which was published in Appended Paper VIII (Kihlman and Lindqvist, 2004).

3.3.1 Experimental Equipment

The method used to perform force analysis was to compare conventional and orbital drilling,
first in a reference case and then in similar tests in the robot. All tests were done on 6mm
thick aircraft aluminum 7475-0 series. The reference measurements were performed using a
Cincinnati NC-machine, where the NC-machine was used for holding and manipulating the
drilling machines and specimens. The conventional and orbital drilling machines were
attached in the NC-machine chuck, and the specimens were attached in the NC-fixture. The
NC-machine could assure a rigid attachment of both specimen and machines. The force
measurements were made with a 3-axis force sensor bed, type Z15063 from Kistler
Instrumente AG, which was originally designed to measure cutting forces in milling. This
sensor bed measures up to 3000N with a resolution of +/- 1N (Bjurstam, 1994). The sample
rate for all tests was 10.000Hz. The coordinate system is defined according to x and y as the
plane and z orthogonal to the plane. The tool changing between the drilling machine and the
other equipment used the Coromant Capto system from SANDVIK. Coromant Capto is
generally used for holding cutting tools in NC-machines, but in this case worked perfectly as
a robot tool changer. This simplified changing between the conventional and orbital drilling
machines, both in the NC-machine and the robot.

52
 Robotic Orbital Drilling

For both the conventional and orbital drilling experiments, the parameters that were varied
were hole diameter and feed rate. For the conventional drilling tests, an Atlas Copco drilling
machine was used with a self-feeding mechanism, which was set to work at 3000 rpm. The
feeding mechanism in the drilling machine used was pneumatic and could not be trusted;
therefore, the feeding movement was performed by the NC-machine and the feed mechanism
in the drilling machine was locked. For the orbital drilling tests, one of the early prototypes of
orbital drilling, shown in Figure 3.7, was used. The spindle was pneumatic and set constant at
12000 rpm. The orbital rotation speed was constant at 120 rpm. The parameters that were
possible to vary were hole diameter and feed rate. Feed rate for both machines alternated
between 0.5, 0.7, 1.0, 1.5, and 2 mm/s. Cutter diameters for the conventional drilling machine
changed between 6, 7 and 7.9 mm. For the orbital drilling experiments, a solid carbide 4 mm
cutter was used and the offset was adjusted to match the hole diameter.

Figure 3.7: The Portable Orbital Drilling Unit (PODU)

Changing the hole diameter of the PODU is a cumbersome operation. The offset cannot be
preset to the desired hole diameter, but it still has to be fine-tuned, hence the machine must
have a drill-in procedure, i.e. where cutting tools and parameters are tested and measured in
order to obtain the final diameter of the hole.

3.3.2 Forces in Conventional Drilling

Conventional drilling can be performed both manually with hand-held drilling machines and
automated using numerically-controlled machines. In manual drilling, templates or drilling
jigs are in most cases necessary to give the cutting tool the right position and orientation. All
manufactured parts have a manufacturing tolerance; this variation affects their placement
within the assembly fixture. The drills are guided by the template which is connected to the
fixture. Therefore, the hole positioning and countersunk holes will be out of round and either
too deep or too shallow proportionally. An attempt to map positional hole accuracy of high-
speed steel twist drills with conventional point geometry was performed in a doctoral
dissertation by Kaminski (1992). Kaminski divided the penetration process in three steps. The
first step occurs at the beginning of the penetration, i.e from the moment the chisel edge
impinges on the surface until it is fully below the workpiece surface and cutting chips of equal
thickness on both its leading halves. The Second step begins where step one ended and lass
until both lands of the drill contact the hole walls. The third step begins when the entire drill
point is engaged, and the outer corners and the lands are in contact with the wall of the hole.
These sections were used to investigate wandering phenomenon, and thus drill deflection
when engaging the surface.

Another thorough investigation of forces and torques in each stage throughout the drilling
process was presented by the University of Wollongong in Australia by Fernandes et al.
(2001). Fernandes describes continued research on thrust force analysis in Fernandes and
Cook (2005), and later a doctoral dissertation (Fernandes, 2005), where the factors affecting
maximum thrust force and torque in conventional drilling of carbon fiber were investigated.
Fernandes and Cook also developed an empirical model based on Shaws simplified equation.
By using this model and a slight modification to Shaws equation, they could take in
prerequisites such as hole diameter, hole thickness, and the number of previously drilled

53
Robotic Orbital Drilling

holes, and by knowing the force signature throughout the drilled hole, they could also
determine that cutting parameters, such as feed and spindle speed, were varied to prolong tool
wear and facilitate hole quality. In the robot laboratory at Linkping University a replicated
experiment to discover the same force signature as was presented in Fernandes et al. (2001).
These experiments were performed using a conventional spiral drill instead of the one-shot
drill used at University of Wollongong.

As shown in Figure 3.8, the cutter has been divided into two sections. The first section is the
conical part of the drill from first engagement of the material to full penetration. The second
section is the rest of the spiral cutting edge.

1
2

Figure 3.8: The cutter was divided into two sections

The thrust force was measured and, repeating the procedure for around 20 holes, it was
concluded that there was a repeating force signature. Figure 3.9, shows a typical result from
one of the measurements. The effect from the three different stages of the cutter could be
identified throughout the drilling process, where the negative Z is the thrust force, x and y are
the forces in the plane.

I II III IV
conv_05

Section 1 enters
material

Section 1 fully
exits material
Section 2 enters
material Section 1 start
exiting material

Figure 3.9: A representative result from conventional drilling with a hole diameter of
6 and feed rate of 0.7mm/s

The concept of analyzing the drilling force signature by dividing the cutting force signature in
stages was taken from Fernandes research, previously presented in this chapter. As seen in
Figure 3.9, stage I shows a linear increase of thrust force from first engaging the material with
the drill tip till section 1 has fully penetrated the material. Stage II shows a more or less
constant force signature as the cutter is fed through the material, while stage III shows a linear
decrease as section 1 exits the material. The obvious question at this point is why forces did
not go back to zero as it theoretically should have as section 1 of the cutter has totally exited
the specimen Based on the theories on exit burrs presented by Nichol (2001), the author has
made the following assumption: by look closer to what happens with the material as the tip of
the cutter is just about to come out on the backside, the force acting on the last millimeter of
aluminum causes the material to plasticize, creating something called a swarf (or alternatively
exit burr). Figure 3.10a shows a schematic view of the event, while Figure 3.10b shows a
photo of what happens with the aluminum as the cutter exits the material.

54
 Robotic Orbital Drilling

swarf

Figure 3.10a: A schematic view of Figure 3.10b: A photo showing what

what happens at the end of stage III happens when the drill
exits the material
The center of the cutter is not machining, but instead is more or less pushed through the
material. This is one reason that the forces are so high in conventional drilling. As the
aluminum is plasticized, the material is pushed forward. If this event did not happen, the
forces would drop to zero at the end of stage III, because the movement on the x-axis in
Figure 3.9 of stage I and stage III should have the same length. In practice, due to the
plasticization of the material, the cutter must machine the material pushed forward throughout
phase IV.

The drilling trials resulted in axial forces between 120 to 320N, depending on feed-rates or
hole diameters. For example, using a 6mm cutting tool in conventional drilling with a feed-
rate of 0.7mm/s, there were axial forces close to 200N. The force in the feed axis (the thrust
force) was derived from the geometry of the cutter in combination with the feed rate. The
forces in the plane are close to zero, with some minor vibrations; this characteristic can be
explained by Figure 3.11. The force equilibrium shows that the resulting forces in x and y
equal zero.

+Y

+X -X +Y
+

-Y +X

Figure 3.11: The force equilibrium for conventional drilling.

In Figure 3.11, the vectors to the right are the coordinate system, which corresponds to the
coordinate system on the Kistler force sensor. Hence, in this case the coordinate system of the
cutter is static in the sensor coordinate system. The small oscillation in Figure 3.9 is
consequently only vibrations.

3.3.3 Forces in Orbital Drilling

The Orbital drilling cutter has two cutting edges. The cutter has full contact on all cutting
edges throughout the machining process, and as the side edge mills, the face side drills, see
Figure 3.12.

55
Robotic Orbital Drilling

Cutting edges on
both side and face

Figure 3.12: A typical orbital drilling cutter

An analogue analysis that was done on the conventional drilling, showing a thrust force
histogram, is presented in Figure 3.13. The orbital drilling process had a sinusoidal force
curve of around zero. These curves correspond to radial cutting forces caused by the orbital
movement in the orbital mechanism, as seen in Figure 3.13. The radial forces oscillated
around +/- 5N independent of both the feed-rates and hole diameters tested. The thrust force
was measured to 30-40N.
The axial edge of
I II III the cutting tool fully
exits material

Entering material

The radial edge of

the cutting tool exits
material
Cutting edge has
orb_05
full contact

Figure 3.13: Orbital drilling with a feed rate of 0.7mm/s,

a 6 mm hole diameter and a 4 mm cutting tool

The radial forces can be theoretically explained as follows: the revolving feed-rate in the
plane causes the chip-thickness to vary in the tool direction. As shown in Figure 3.14, forces
are measured in a fixed coordinate system, whereas the force axis in x and y will vary. This
phenomenon can be compared to the characteristics in a milling process.
+Y
+X

+
-X

+Y -Y

+X -X
+
+Y
-Y

feed-rate(x,y) +X

Figure 3.14: Force equilibrium in orbital drilling

(coordinate system to the right)

56
 Robotic Orbital Drilling

The orbital drilling process had a sinusoidal force curve of around zero. These curves
correspond to radial cutting forces caused by the orbital movement in the orbital mechanism.

3.3.4 Force Comparison

There was a clear trend that the forces in orbital drilling did not change when the feed rate
was increased, as seen in Figure 3.15b. The conventional drilling has a linear increase with
increasing feed rate, see Figure 3.15a.

Figure 3.15a: The proportional Figure 3.15b: The thrust force

increase of thrust force when in relation to increased feed rate
increasing the feed rate in in orbital drilling
conventional drilling
The experiments showed that forces in orbital drilling can be as much as 8 times lower than
conventional drilling. This trend becomes more evident when the feed-rate or hole diameter is
increased. Interestingly, the axial force did not rise particularly much due to either increased
feed-rate or larger hole diameter in orbital drilling. This can be explained by the fact that the
feed rate is increased when more material is machined radially in proportion to the axial
machining, and hence a low proportional increase in thrust force.

3.4 Robot Orbital Drilling

The next phase of the experiments was to investigate what happened when the PODU was
attached as a tool to the robot. This would be the first time robotic orbital drilling was
undertaken, and was done in two stages in the ADFAST project as part of this research. The
first experiments were done by the author in the robot lab at Linkping University, using the
PODU; the second test was done at the BAE SYSTEMS research lab in Bristol, UK, using the
CNC ODU.

3.4.1 The Preliminary Test Bed Using The PODU

The same force sensor as in the reference measurements was used with the robot, which was
attached to a stiff frame. The robot used for the tests was an IRB4400, 60 kg payload robot
from ABB with an S4Cplus controller. Figure 3.16 shows PODU attached to the robot.

57
Robotic Orbital Drilling

Figure 3.16: The PODU attached to the robot

Force measurement in robotic orbital drilling

Force measurements were done only to verify the machining process. At this stage,
experiments were made using only the orbital drilling unit, and not the conventional drilling
machine. The goal with the robot drilling was to investigate if orbital drilling was possible in
a standard industrial robot; therefore, no conventional drilling was done using the robot. It
was concluded that force measurements in this experiment showed similar behavior as in the
measurements in the Cincinnati machine. Forces in the axial direction were only slightly
lower when drilling with the robot compared with the reference measurement.

Deflection measurement in robotic orbital drilling

The most important test in the robotic orbital drilling experiments was to measure deflection
of the robot during drilling. Since robotic orbital drilling had never been done before, it was
decided to start by having the orbital drilling unit hanging freely in the robot. This meant
that there was neither a pressure foot used nor a drilling jig to secure the machine; the only
thing in contact with the workpiece was the cutter.

Dislocation was measured with three LVDT sensors (Linear Variable Differential
Transformer) as shown in Figure 3.17a. These sensors had a resolution of 0.001mm and a
range of +/-2.5 mm. The deflection was measured in the x, y and z axes, where x is the axial
direction, y is the horizontal direction and z is the vertical direction.

58
 Robotic Orbital Drilling

X
Y
Z

Figure 3.17a: The LVDT-sensors Figure 3.17b: The LVDT-rig

The L-profile was installed to the front end of the end-effector. The result from the deflection
measurements when the end-effector is hanging freely is shown in Figure 3.18.

mm
Z

Sample (10.000Hz)

Figure 3.18: Typical appearance of deflection in the robot when the orbital drilling
unit is held freely in the robot TCP

The axial movement started out with an offset of 0.06mm, followed by a sinusoidal
movement with an oscillation of +/-0.005mm. This was derived from the fact that the forces
detected from the reference measurement clearly stated that the axial forces were greater than
the radial forces. However, having the end-effector held freely resulted in a rather large
horizontal and vertical movement compared to the relationship between the forces in the
reference measurements. The vertical movement started with an offset of 0.03mm, followed
by an oscillation of +/-0.01mm. This can be explained from the fifth axis of the robot that
deflected more than the others. This is a result of having the end-effector held horizontally,
resulting in a distance between the axial force and the fifth axis rotation point. The deflection
in the horizontal direction gave +/-0.02mm. This is more logical when there is symmetry on
both sides of the end-effector.

These preliminary experiments clearly show that the deflection during drilling as excessive,
with the conclusion being that the end-effector must be pressed to the surface to hinder the
movement in the plane. One approach to solve this solution was to develop a pressure foot
between the specimen and the PODU unit. The solution was called the Linkping Rubber
Press and Vacuum Suction (LRPVS). A rubber ring was installed as a high friction material to

59
Robotic Orbital Drilling

avoid oscillation, as seen in Figure 3.19. The unit was pushed with the robot against the
workpiece, where the friction stiffened up the drilling machine.
Vacuum boost
attachment

Rubber
ring Vacuum

Figure 3.19: A picture of the LRPVS concept

The same deflection measurements were done using the LVDT-sensors. Figure 3.20 shows
the deflection during the drilling cycle, while Figure 3.21 shows the forces during the same
drilling cycle.

mm
X

Z
Y

Orb_d_05

Figure 3.20: Deflection measurements in the robot using the LRPVS-concept

60
 Robotic Orbital Drilling

pre-force

Figure 3.21: Force measured with the LRPVS concept with 0.5mm/s, 6 mm hole
diameter and a 4 mm cutting tool

Note that the force in the z-axis in Figure 3.21 has a constant offset. This force corresponds to
the pre-force caused by the robot. The movement in z (vertical direction) became close to
zero; this meant that the LRPVS reduced some of the deflections. However, in the y direction
(horizontal) there was still some movement. An explanation for this is that the rubber ring was
too thick. In order to reduce the thickness of the rubber, a flange in the orbital drill unit had to
be removed. In an optimal application, the rubber thickness should be as thin as possible to
reduce the inherent movement of the rubber. The movement of 10m in y could have been
zeroed out if the rubber had been thinner, and if the pre-force (robot pressing against the
plate) had been optimized. The reason for the movement in the x direction (axial direction)
comes from not having trimmed the pre-force. The pre-force must be at least equal to the axial
force arising in the drilling process, which was not the case in the tests above; this will require
further experiments. Further studies in this area, presented in Paper IV (Kihlman et al., 2002),
showed that a metallic ring had enough friction to avoid oscillation. This method was further
employed during the implementation of the CNC ODU.

3.4.2 The Final Test Bed Using the CNC Orbital Drilling Unit
Later in the project, a larger demonstrator was built at Sowerby Research Center at BAE
SYSTEMS in Filton UK during February 2004. This work was undertaken by personnel at
Novator AB and BAE SYSTEMS, and managed by Airbus. The author of this dissertation
had only a role in this part of the research: to analyze the result from that installation. Indeed,
in a project like ADFAST, exemplary teamwork was always the case, where all participants
cooperated in the formulation of an implementation like this. The actual physical work,
however, should be credited to the three companies mentioned above. This section will
summarize these results.

Experimental Setup
The demonstration was held in order to answer to the call within the ADFAST project for a
practical demonstration of the CNC orbital drilling unit mounted into a robot cell. The demo
cell was comprised of a CNC orbital drilling unit, a TwinSpinTM 102 attached, a KUKA KR
200, a 200 kg payload robot and a fixture holding the specimens. The samples drilled in the
experiments were aerospace CFC (Carbon Reinforced Composites) and Aluminum. This
dissertation will only present the results on the Aluminum specimen. The specimen was 13
mm thick and the holes drilled were inch in diameter. The drilling was done in two tempos
first a rough cut, followed by a fine cut. The rough cut was performed with a feed rate of
600 mm/min, an orbital speed of 400 rpm and a spindle speed of 32670 rpm. The fine cut was
performed with a feed rate of 400 mm/min, an orbital speed 400 rpm, a spindle speed of
32670 rpm and an increase of eccentricity of 0.15 mm.

61
Robotic Orbital Drilling

The CNC ODU End-effector

One of the drilling work package objectives within the ADFAST project, apart from actually
building the CNC orbital drilling unit, was to ensure that the unit would be mass-balanced in
operation. Mass-balanced means that although the spindle inside the orbital drilling unit
revolves in a spectrum from 0-500 rpm with a diameter offset of 0-10 mm, most of the
subsequent oscillating force is eliminated. It is important to have in mind that the spindle has
a high-speed, rotating between 0-40,000 rpm. Tests performed at Novator AB before delivery
to BAE SYSTEMS indicated no more than 10 N in total. The end-effector unit, including
vacuum foot and mounting adapter plate, had a total weight of 150 kg, where the CNC ODU
was 105 kg.

Experiences from earlier experiments on the PODU suggest that a pressure foot is important
even though the radial forces are small. In this installation, no rubber was used. Instead, a
metal ring, surrounding the cutter, was used as a pressure foot to engage the surface. A
guiding rail mechanism made the unit slide and push against a spring. As the spring became
pressurized to a certain force, an inductive position sensor was indicated to stop the robot
from pushing the unit harder to the surface. Unfortunately, a jamming effect in the mechanism
hindered the feeding mechanism from working smoothly, and a temporary solution had to be
made. The jamming effect arose from the offset distance from the center of the metal ring to
the pressurized location, as is seen in Figure 3.22. The jamming effect was eliminated by
installing a plastic block closer to the rail mechanism to reduce the lever effect between the
metallic ring and the guiding rail.

Guiding rail

Plastic block

Metallic ring

Figure 3.22: The CNC-ODU End-effector

If more time were available, the jamming effect had been eliminated and the metallic ring
would have been used. The metallic ring is a way to make the tip of the end-effector slimmer.
In this demonstrator, however, that was not an issue, since there was open access to the
specimen. An adjustable spring adjusted the pre-pressure force.

Hole Quality Analysis

For the hole quality measurements, a Mitutoyo digital probe was used, where the largest and
smallest diameter of each hole was inspected to detect ovality etc. A Mitutoyo surface test
unit was also used for surface finish measurements. All these tests are presented in the
62
 Robotic Orbital Drilling

Appended Paper VIII (Lindqvist and Kihlman, 2004). The hole quality measurements and
surface finish measurements were performed by Novator AB at their lab in Stockholm.

The results from the measured data gave H8 holes (0,027 mm), a hole roundness of 0,013 and
a surface finish below 0,68 Ra m, which is well below requirements (for example Lockbolts
of Ra 1,6 m). Drilling H8 holes was one of the most important results for Novator in the
ADFAST project, so these results proved well for the requirements.

3.5 Final Remarks on Robotic Orbital Drilling

It was concluded that orbital drilling reduces the thrust force 8-10 times when compared to
conventional drilling. Orbital drilling has a radial force below 10 N that can be taken up by a
pressure foot. The pressure foot is preferably made of metal, as rubber causes small internal
movements. The test showed that the metal ring did not damage the material whether CFC or
aluminum was used. At the time this dissertation was published, new end-effectors which
weigh less and yet still use numeric control to change spindle speed, feed rate and offset are
being developed. The input to this research can conclude that the PODU is preferred if the
circumstances are constant and the robot manage only smaller payload. Due to the weight is
high for the CNC-ODU used in this research a larger robot is required. The advantage with
the use of the CNC-ODU is its ability to change cutting data continuously during the drilling
process. Linking the results in this chapter to the conclusions from Chapter 2 indicate that
orbital drilling is an enabling technology for drilling applications in the aerospace industry,
which is primarily due to the low thrust forces. Even if the thrust forces in orbital drilling are
small a pre-loaded pressure foot is still required. The authors experience tells that even the
smallest forces exerted on the TCP will cause movements. This is caused by backlash (i.e.
play) in the robot, especially if there is an oscillating force. In the robot lab, there have been
several experiments conducted to investigate this matter. Some of these results were later
presented by Sunnanbo (2003), a subproject supervised by the author. Moreover, Eriksson
Helle (2002), another subproject supervised by the author, concluded that even larger forces
makes the robot to act like a spring. Eriksson Helle (2002) presented a spring model for this
phenomenon. The typical approach to deal with this problem is to model the environment for
the robot in order to predict the phenomena, such as was discussed in Section 2.6.2; however,
this will only solve the static scenario. Orbital drilling showed that if only a small pre-force
was applied, the backlash was removed and the deviation of the robot was small enough to
provide holes within aerospace requirements, as was presented in Section 3.4.2.

The author, together with Mr. Van Duin, during the authors stay at the University of
Wollongong in Australia in 2004, developed a solution to avoid all these worries. The Force
Cone Concept was developed to detect any deviation as the pre-force is applied. Using a cone
between the drilling machine and the airframe to be drilled is a common approach to suck out
the chips and dust produced in the drilling. This is especially important when drilling carbon
fiber since the dust is hazardous. The concept, designed and tested by the author and Mr. Van
Duin, was based on using three force sensors, called load cells between the force cone and the
drilling machine. The three load cells positioned on an equivalent angle on a circle detected
the dis-normality between the cone and the material. Figure 3.23a shows the cone on a
principal drilling machine, while Figure 3.23b shows how the force signature moves closer to
the bulls-eye of the XY- plane. Each ring in the polar plot in Figure 3.23b is force magnitude
and the quadrant location show the direction of the force vector. When the force reaches the
center of the plot, all sensors have equal value, which was assumed to be equivalent with
perfect normality. An additional photo on the physical Force Cone Concept is shown in
Appendix D.

63
Robotic Orbital Drilling

Load cell 1
Force
2 Load cells
measurements
+
+
+ +
Cutter +

Load cell 3

Cone Load cell 2

Spindle
Figure 3.23b: The XY-
force signature, adjusting
Figure 3.23a: Cross
the robot in 5 iterations
section of the force cone
The force cone concept is an easy and cheap solution to make sure the robot stays normal to
the surface prior to drilling. Moreover, the load cells are not only used to measure normality;
they are also used to measure the total preload on the surface. The total preload is the sum of
the forces in all three load cells. One convenient scenario for this concept is flexible materials.
If the material bends as the preload is engaged, it is normally difficult to compensate for a loss
of normality. The force cone concept will notice this phenomenon and compensate
accordingly. The research between the author and Mr. Van Duin will be presented at next
SAE conference.

Furthermore, the low cutting forces in the orbital drilling process itself may be the solution to
make stack drilling without deburring, as many tests that have been performed in this research
which indicate that orbital drilling eliminates deburring in many applications. The cost model
presented in this chapter proved that orbital drilling can reduce cycle time by one-half, much
due to that the hole can be drilled in one attempt. The cost model study also states that the
cost compared to conventional drilling can, depending on the application, be reduced by 50%,
especially for machining, consumables and temporary fastening.

Small thrust forces in orbital drilling have another important context in the drilling composite
laminates. High trust forces may cause composite laminates to delaminate during drilling. In
the extreme case, the SAAB Ericsson Space Bread Board application proved that drilling a
0.8mm thick composite laminate did not delaminate in orbital drilling, which alone was a
great challenge.

One of the remaining challenges with orbital drilling is to enable a faster hole diameter
calibration procedure. In conventional drilling, when a cutter is changed the hole diameter is
essentially the size of the drill. With orbital drilling, the eccentricity, constituting the size of
the hole diameter, must be trimmed by drilling holes and measuring the hole diameters, which
is cumbersome. One method speculated by the author could be to measure the eccentricity
using a non-contact sensor (e.g. inductive sensor or laser beam). Measuring eccentricity,
while driving the machine, would indicate to the operator how to adjust manually, or be
performed automatically in a numerical machine.

64
 Affordable Reconfigurable Tooling

4 Affordable Reconfigurable Tooling

Chapter 3 presented Orbital drilling - a new technology that eliminate some of the problems
in using industrial robots for drilling. Assuming now that industrial robots are out on the
workshop floor for drilling automation, Chapter 4 will show how robots can be utilized more,
by using them to actually perform changeover between different products. Chapter 4 is
presenting a new tooling technology that utilizes the same industrial robots used for drilling
automation to perform the actual reconfiguration.

4.1 Aircraft Tooling Theory

Even if the Swedish inventor Christopher Polhem was among the first to use the concept of
interchangeable parts in the manufacturing of clock gears at the beginning of the 1700s, it was
Eli Whitney, who started using machines for the manufacturing of muskets in the end of the
same century, that removed the filing operation in the final stages of assembly. However,
using jigs and fixtures in a large scale was introduced by the pioneer of automotive
manufacturing, Henry Ford (Hopp and Spearman, 1999). Mr. Ford used tooling as holding
devices to rapidly build cars using interchangeable parts.

The system used for positioning parts in space during assembly is known as tooling, which
can be divided into two different principal groups: fixtures and jigs. Fixtures position and
hold parts during assembly, whereas jigs not only position and hold parts during assembly,
but are also used to guide cutting tools (Hoffman, 1984; Pollack, 1976). The most common
case where jigs are used in aircraft assembly is in drilling jigs. The drilling jig is normally
attached to the framework in front of an aircraft panel, while the cutting tool is both
positioned and guided through a bushing installed in the drill jig. The research on tooling
presented in this dissertation is mainly focused on assembly fixtures. Tools are needed to hold
a workpiece in an assigned position. In this position, a variety of tasks may be performed,
such as turning, milling, drilling and grinding (Kakish et al., 2000).

Fixating a workpiece under various circumstances, there are a total of 12 (2 x 3 x 2) linear and
rotational movements along the x, y and z axes, including both positive and negative
directions (Trappey and Liu, 1990). The way of confine a workpiece in six degrees of
freedom is as follows: three fixed points for supporting the first surface, two fixed points for
the second surface, and one fixed point for the third surface; hence, this is called the 3-2-1
method (Pollack, 1976).

4.1.1 The Generic Tooling Model

Although there are an astronomical number of tooling designs and methods for holding parts
during assembly, in this section a generic tooling model for aircraft assembly tooling is
presented. It has been designed by the author as a method to help consider important aspects
regarding the design of a tool, and will be used in this chapter to compare existing tooling
technologies with the concept presented in Section 4.3 (see Figure 4.1). In this approach, the
decision has been made to work from the workpiece and outwards to the surroundings. One
should be aware that it could as well be possible to work inwards, starting with a surrounding
system, and work inwards to the workpiece.

65
Affordable Reconfigurable Tooling

Zone IV Surrounding system

Zone III Distance supports

Zone II
Pick-ups
Zone I

Workpiece

Figure 4.1: A generic tooling model of an aircraft assembly fixture

Zone I The Workpiece

The generic tooling model is divided in different zones (see Figure 4.1). Zone I is the
workpiece. Normally when beginning to design a new tooling concept, the tool engineer
studies the product, i.e. the airframe, and especially the tolerance requirements for the
product. Decisions are made regarding which sections of the parts need to be positioned with
higher accuracy than other sections. The locations of the workpiece that require higher
demands on accuracy are called datum points, or just datums. One typical example of a
datum point is one of the points that need to be located in a hinge line. Building a hinge line
in one factory and shipping it to a final assembly somewhere else in the world demands the
hinge to fit with its mating part without jamming, which can be quite challenging.

Zone II The Pick-ups

Once the datums are clarified, the engineer can start studying how to design a locator, called a
pick-up. This feature picks up the workpiece, positions it to a point within the required
tolerance, and maintains that tolerance throughout the assembly process. This is challenging if
the point being located is inside a complex structure, or if the product weight or dynamic
process forces are high. The pick-up in Figure 4.1 corresponds to Zone II in the generic
tooling model. A pick-up is often a standardized clamp device or a piece of metal with a hole
that holds the product with a bolt.

Zone III The Distance Supports

The location of the pick-up must be supported by something which can hold the pick-up itself
steady in space. Pick-ups are small features whose only purpose is to position and hold,
whereas the distance support holds the actual pick-up. The distance support, shown in Figure
4.1, corresponds to Zone III. The distance support can be designed in many different ways,
but it is normally in the size range of around 0.5 m. Later in this chapter, some of the many
different names for the distance supports in different fixture technologies for different tooling
technologies will be presented. It is the distance support and pick-up together that enable
complex airframe parts to be attached in the tooling.

Zone IV - Surrounding System

Holding the Distance support in space is done by the surrounding system. In Figure 4.1, the
surrounding system corresponds to Zone IV. The surrounding system is in most cases a
framework beams that together constitute a cage where the pick-ups can interface rigidly.
But in some tooling technologies, the surrounding holding the pick-ups is comprised of a
foundation on the floor or on the wall. It does not matter if the surrounding system is a
foundation on the floor or a framework; the common goal for the surrounding system is to
rigidly hold the Distance supports throughout the assembly process.
66
 Affordable Reconfigurable Tooling

4.2 Existing Tooling Solutions

Before introducing the new tooling solution which has been developed in this research, the
following section will present a short summary of existing tooling technology. These existing
solutions have been used as reference points to the new tooling solution presented later in the
chapter. By combining advantages from existing solutions, a first approach for the new
concept was formed.

4.2.1 Conventional Tooling

The most common tooling technology for airframe assembly used today is conventional
tooling (CT). The surrounding system in CT consists of steel beams that are welded together.
CT uses distance supports and pick-ups that are tailored for a specific purpose. For example,
one CT is manufactured for the left aileron aircraft structure, while another for the right
aileron structure. The pick-ups, however, can in many cases be standardized. Final accuracy
of a pick-up is achieved using a metrology system, such as cameras or lasers. Metrology
systems are presented in Chapter 5.

With the use of a fluent resin called shimming, distancing the distance supports (Zone III in
Figure 4.1) from the framework slightly to take up un-tolerances is accomplished. The most
common material used in conventional tooling is standard steel, although there are occasions
when aluminum is used, such as when there are requirements for low weight, for example in
turning fixtures or mobile fixtures. It is important to have in mind when selecting material that
steel is three times stiffer, and has three times less thermal expansion, than aluminum.

Due to the fact that CT is tailored to a specific application, each assembly normally has its
own dedicated tool. That is why CT also goes under the name dedicated tooling, or sometimes
product-specific tooling. In addition, when building a complex product such as an aircraft, the
final design is hampered by changes that immediately affect the tooling design. This will lead
to a cumbersome operation to modify the welded or shimmed pick-ups to new locations.
Thus, the lead time for using conventional tooling is long, often more than six months.

CT is a well-known, safe method which follows the old traditions of aircraft manufacturing. It
is also over-constrained to maintain accuracy, can be tailored to any application and it follows
the law of the 3-2-1 method (discussed in Section 4.1). As mentioned in Chapter 1,
Monumental Machines have a limited ability to handle changeovers. Conventional tooling
cannot be reconfigured to another product type at all; hence, it can be referred as a kind of
Monumental Tooling.

4.2.2 Modular Tooling

Conventional tooling is a solution that was tailored to one product type. Modular tooling
(MT) is a tooling solution that is built for a dedicated purpose as well, but the surrounding
system and distance supports are designed from a toolbox of modular components. The
framework in CT is welded together, whereas in MT, the framework is screwed together. This
is possible by using the slots created in the extrusion process in manufacturing the beams for
the modular tooling framework, as shown in Figures 4.2a and 4.2b.

67
Affordable Reconfigurable Tooling

Figure 4.2a: A Modular tool Figure 4.2b: The building

with one clamping Pick-up concept of Modular tooling

The same slot system is used to attach the distance supports or pick-ups directly to the
surrounding system. The components constituting the tooling can be bought as standard off-
the-shelf items from suppliers, and thus shorten lead times while ensuring a given level of
quality. Even if modular tooling is exposed for larger thermal expansion and requires larger
dimensions, it has proved to be a cost-effective technology to use as assembly fixtures for
aircraft manufacturing.

Screwing the beams together requires a different building method when building the modular
tooling compared to conventional tooling. It is convenient to use a laser tracker such as that
presented in Chapter 5. This method requires experienced personnel, but still has proven to
shorten lead times compared to conventional tooling. BAE SYSTEMS has a patented build
method for modular tooling (Lindgren, 2001).

Even if changes are easier to perform compared to conventional tooling, they are still
dedicated to one product and are rarely reconfigured for a changeover. However, modular
tooling enables re-use of tooling components to recycle the tooling for future tools, hence
modular tooling is not likely to be reconfigured between products, but can be rebuilt. The idea
of rebuild and modularity itself has made a great contribution to the ideas of the new tooling
technology presented in Section 4.3.

4.2.3 CNC-controlled Tooling

CNC-controlled tooling is a suitable solution when the product volumes are low and the the
tooling is reconfigured often. The distance supports of the CNC-controlled tooling have built
in encoders and servos to become actuators instead of static structures. These actuators are
commonly known as Pogo-sticks, or Pogos. (Kostyrka and Kowalsky, 2000; Munk, C., 2002;
Marin et al., 2002). One numerical controller handles the movement of all Pogo-sticks. The
surrounding system is a rigid foundation, and the pick-ups are in most cases vacuum cups.
The most common surrounding system solution is to install the a matrix of Pogo-sticks.
Sometimes this solution is referred to as a bed of nails (Figure 4.3a) or a wall of nails
(Figure 4.3b), depending on if the Pogo-stick bed is laying on the floor or installed vertically.

68
 Affordable Reconfigurable Tooling

Figure 4.3a: A bed-of-nails with Figure 4.3b: A wall-of-nails with Pogo

Pogo Sticks holding a panel during Sticks holding a panel (source: Vertical
edge machining (source: 3 Axes Torrestool, MTorres)
Horizontal Torrestool, MTorres)

A common commercial application for CNC-controlled tooling using Pogo-ticks is to fixating

aircraft skins during machining operations, such as edge milling and in some cases drilling.
There are exceptions, however. In a paper presented by Marin et al. (2002), Pogo-sticks were
used for holding aileron structures for the Airbus A320 A330/340 and A340-500/600 aircraft
families while drilling and installing different solid rivets in the trailing edge of the elevators.
Nail-configuration for these kinds of tools is common today, especially in the aerospace
industry.

Other advanced CNC-controlled tools that went beyond just nail configurations have been
undertaken in aerospace projects. One major accomplishment was made at Boeing in the early
1990s (Olsen, 1990; Wright and Sarh, 1989; Sarh, 1992; Sarh, 2002). Boeing developed a
highly-automated cell that used a multi-robotic system with several exchangeable multi-
functional end-effectors and computer-controlled flexible fixtures. The system could assemble
first-level aircraft structures (ribs, bulkheads) including part recognition, pick-up, placement,
tackling, and riveting. In effect, the system could replace conventional tooling and be able to
quickly reconfigure itself for new types of subassemblies. Another similar approach was
undertaken by Li et al. (1996); this group was able to conceptualize a flexible workcell for
aircraft wing spar assembly. The fixture included 76 step motors controlled by a central
controller. The concept could also handle different types of structures as well as left and right
wing assembly for a single aircraft.

There have been similar approaches in the automotive industry as well. One impressive
installation was made at Nissan Motor Co., Ltd (Naitoh et al., 1993). Nissan had problems
with handling 20 new car models every year, as each model changeover required two and a
half years production lead time. By implementing a programmable fixture with 252 servo
axes to control 35 positioning robots and 16 welding robots, Nissan developed a system that
could, in theory, produce any of its vehicle models. Through this system, Nissan managed a
product output of 20 thousand cars per month, saving 80% of the cost per each changeover.
This system enabled 8 types of vehicles to be produced on a single line, compared with 2
types when using conventional methods.

One problem with CNC-controlled tooling is the high investment cost. There has been,
however, one attempt to design a manually reconfigured bed-of-nails tool to lower the
investment cost and complexity. Sela et al. (1997) presented a project on a novel jointed
reconfigurable modular system for the fixturing of thin-walled flexible objects, using the idea
of the t-slot modular technique to attach pogo sticks on a base plate. An operator could then
manipulate the Pogo-sticks by screwing a height adjustment knob to set the length of the
69
Affordable Reconfigurable Tooling

stick. By doing so, the group was able to develop a pogo bed that was reconfigurable in 3
axes. They also presented a parametric methodology to find the shape and location of the
support wall to the workpiece.

4.2.4 The Need for New Tooling Technology

Before presenting this researchs new tooling technology, Figure 4.4 is shown to help position
the gap in todays tooling technology. This gap can be thought of as where this research took
its first step towards inventing something that did not exist.

Setup time
Conventional tooling
Modular tooling
month
Tooling in this research
weeks
CNC-controlled tooling
hours

minutes
Changeover
never 1/year 1/week 1/day frequency

Figure 4.4: A diagram pointing out the gap of assembly tooling technology

This gap of technology was first discussed by Saab Aerostructures and presented by Engstrm
(1998). The solution was developed in a project founded by the European Union, called
Flexible Workshop for Airframe Assembly (FWA), developed the concept, with the result
being a RobCad simulation showing the idea. The project generated a fixture that was
possible to configure in three axes using a robot manipulator. The fixture consisted of a wall
of Distance supports where airframe panels could be mounted. The study showed that
different side panels on a SAAB 340 aircraft could be assembled. The idea was later patented
by SAAB Aerospace (Saab Aerostructures, 2004). The research presented in this dissertation
was carried on the FWA project at Saab Aerostructures, to what later became the fully
functional physical demonstrator presented in Section 4.4.

Starting at the top left in the figure, it is seen how conventional tooling is a technology that is
built to suit one product type; hence, these fixtures and jigs are not reused, rebuilt or
reconfigured for the next generation of tooling. Modular tooling, on the other hand, has the
ability to be re-used and rebuilt for the next generation of tooling, but not reconfigured in a
changeover. Modular tooling is still dedicated to one assembly. CNC-controlled tooling is the
existing tooling technology that is designed for handling changeovers using one tooling
system. To avoid CNC-controlled tooling becoming Monumental, the tooling requires a
production scenario where it is reconfigured often, maybe once per day or more. There did not
exist any tooling solution that was suitable for reconfiguration once a week.

There did not exist a tool that could be reconfigured once a week with a setup time in a few
hours. Figure 4.4 show a dashed circle to point out this gap. The vision was to be able to
reconfigure a fixture in a matter of hours for a limited geometrical range. In addition, there
was a need to enable the tooling to be rebuilt if the product geometry did vary more. It is
common to think that flexibility is expensive. In the following section, a new tooling
technology will be presented where an assembly tool can be reconfigured within a product
family and rebuilt between product families yet still at a low cost.
70
 Affordable Reconfigurable Tooling

4.3 Affordable Reconfigurable Tooling

The concept of Affordable Reconfigurable Tooling (ART) grew from extensive product
development efforts in the ADFAST team. A preliminary ART concept was presented in
Paper I, and further developed in Paper II; the author of this dissertation also presented a
Licentiate Thesis on the subject (Kihlman, 2002). The concept was based on using a standard
industrial robot, already used to automate drilling, to perform the movement of flexible
tooling modules similar to the Pogo-sticks presented in Section 4.2.3. In this research, these
are referred to as Dynamic Modules (DM). Similar to the CNC-controlled tooling, DMs can
be reconfigured, but instead of having built-in servos and encoders in the DMs, the movement
is caused by the robot manipulator. The DMs are attached to a surrounding system called the
Modular Framework (MF). The MF is not a welded framework, but is bolted together using a
modular construction kit called Box-joints. By using bolts in the Box-joints, the Modular
Framework can be rebuilt similarly to how beams are attached in modular tooling. The
accuracy in positioning the DMs using the robot manipulator is achieved by integrating a
metrology system; the integration between the metrology system and the robot is presented in
Chapter 5. This dissertation takes the conceptual idea presented by the author in Kihlman
(2002) to further present the operation planning and the physical evaluation aspects. Figure
4.5 shows an overview of the system.

Dynamic
Modular Modules
Framework

Metrology Box-joints
system

Figure 4.5: A system overview of ART

The Modular Framework (Zone IV in Figure 4.1), Dynamic Modules (Zone III in Figure 4.1)
and pick-ups (Zone II in Figure 4.1) are presented later in this chapter. Note that the appended
papers I, II and III use the name Static Framework for describing the Modular Framework.
The author has modified that name in this dissertation to create a better meaning. The
framework is not static in this context; hence, modular is a more suitable word to meet its
conceptual functionality. Each subsequent section in this chapter will present the concept of
each sub-system, and also give insight to their physical implementation.

4.3.1 Modular Framework

The approach for the Surrounding System of the ART concept was to use a rigid framework
structure. Therefore, the decision was made to use a steel material instead of aluminum.
Lindgren (2001) indicated that extensive Finite Element Analysis (FEA) was necessary if one
were to employ aluminum as the material choice for the beams in the framework. At this time
the ADFAST team bringing about the ART solution decided not to take any chances and thus
lose accuracy from movement in the framework. Therefore, steel was selected for the
Modular Framework. It was also decided that the aerospace industry should not falsify our
approach from something as trivial as beams. In addition, conventional tooling, which still

71
Affordable Reconfigurable Tooling

was the most common tooling technology, was over-dimensioned to a great extent, and not
surprisingly large steel beams were the only way to convince people that this would work. In
these discussions, an important innovation that would prove to be a breakthrough for the
project was born. Welding the beams would eliminate the possibility to rebuild, as could be
done with modular tooling, and using aluminum was out of the question. The idea that arose
was this: why not design a box that attaches beams to one another? From these discussions
came the Box-joint concept, illustrated in Figure 4.6. The father of the Box-joint concept is
Gilbert Ossbahr, a senior scientist at the department of Mechanical Engineering at Linkping
University.

Figure 4.6: The Box-joint concept

The Box-joint concept is a straightforward approach to attaching beams to each other without
welds. Fixing plates are attached to each side of the beam and bolted to each other to form
fixing devices using standard M14 bolts. The Box-joint concept enables seamless adjustment
of the beams, i.e. no holes on the beams need to be drilled. In addition, the Box-joints are
used for many different purposes. The primary function for the Box-joint is to attach beams
(see Figure 4.7b), but the secondary purpose is to attach the Dynamic Modules to the beams.
The Box-joint concept is what enables rebuild of the framework. Figure 4.7a shows two
beams joined by a fixing device according to the Box-joint principle.

Figure 4.7a: Two beams joined

Figure 4.7b: The Box joint
by a Box-joint construction kit

72
 Affordable Reconfigurable Tooling

The Box-joints are designed to facilitate dimension changes and is as a kind of construction
kit. Going from a large size beam to a smaller dimension is done by this system. Figure 4.7b
shows the modular concept of fixing plates forming the construction kit. At the time of the
implementation, the Box-joint system was not possible to
purchase off-the-shelf, and thus had to be manufactured at
Linkping University. One innovation regarding the Box-joints F
are the nuts used to attach the screws. The holes where the nuts
are installed are made conical with tight fittings in order to avoid
the nuts from slipping as the screws are tightened. A more
thorough presentation of Box-joint and finite element analysis
on the Modular Framework is presented in Pradeau (2004).
Pradeau (2004) concluded that for the Box-joints, there is a
linear relationship between load and deflection under low forces. Figure 4.8: The
This particular conclusion was based on applying load according stiffness experiment
to Figure 4.8. Pradeau concluded that due to the slipperiness of
the protective coating on the beams, the stiffness drops if the load is very high. This can be
explained from the varnish used to protect the beams from corrosion that has a low friction
coefficient. For the purpose of the demonstrator, this phenomenon did not affect the results.

4.3.2 Dynamic Modules

The Modular Framework enables rebuild. That means that the framework can be dismantled,
or loosened up and changed in order to permit a large geometrical changeover, which in
turn facilitates the rebuild of the framework for another product family. If the geometrical
changeover is not too large, but within the range of decimeters, reconfiguration is performed
to change the configuration of the ART. The reconfiguration is enabled using Dynamic
Modules. A Dynamic Module is a passive distance support that is configured and
reconfigured by the robot manipulator. The Dynamic Modules presented in this dissertation
are all flexible in 6 degrees of freedom (6DOF). Numerous concepts were on the drawing
board that had flexibility in the full range from 1DOF to 6DOF. From preliminary tests it was
soon realized that the robots lack of accuracy, especially in moving on straight lines, could
cause high tension in the fixture modules, or even cause the robot to end up in a collision
warning (Eriksson Helle, 2000). Therefore, it was decided in this stage that the fixture
modules reconfigured by the robot had to be flexible in six degrees of freedom (6DOF).

The philosophy in designing a DM is to think modularly! The DM should constitute a

combination of a range of different sub-modules. This means that, depending on the purpose
for using a DM, the design should be possible to change within the range of sub-modules. For
example, a ball-joint module giving 3DOF of rotations can be used for any one of the serial-
linked mechanisms, where a serial-linked mechanism is one of the conceptual DM solutions.
The other conceptual DM solution is the parallel mechanical DM. All concepts of DMs are
presented later on in this section. There is a range of different concepts of DMs. Basically, a
typical company using this technology would specialize in two different DM solutions, i.e.
one serial-linked DM and one parallel mechanic DM, in order to undertake the full spectrum
of force scenarios in the ART system. In this research, as many solutions as were
economically feasible were developed and evaluated in order to learn as much as possible
about this new tooling technology.

The Hydro-mechanical Locking Sleeves

DMs should not be confused with parallel kinematic robots, since the DMs are passive. This
means that the DM is configured by the robot, and as the robot releases (undocks) a DM, the
DM maintains that configuration until next time the robot is docked to it. Maintaining the

73
Affordable Reconfigurable Tooling

configuration is done using a locking mechanism. One challenge in this development was to
try to achieve a locking mechanism that could stay locked without having pneumatic or
hydraulic pressure applied. End-users in the ADFAST project declared the risks and problems
of an actively pressurized system for a long period of time, partly from an injury perspective
but also from the perspective of what would happen if the pressure accidentally dropped. ETP
Transmission, a Swedish company, was a sub-contractor of chuck systems. This chuck system
used a hydro-mechanical locking method. By moving a wedge in the axial direction from a
hydraulic pressure, the wedge clamped up in radial direction, causing a gripping effect on the
cutter as seen in Figure 4.9.

Figure 4.9: The hydro-mechanical wedge locking sleeve

The wedge concept also maintained a locked position until hydraulic pressure was introduced
from the other side of the wedge. This concept was applied to the DMs for locking the joints
in both the serial and parallel mechanical DMs. One locking sleeve early in the project was
built using a 350/250 bar hydro-mechanical locking sleeve, the ETP-HYPUS 20. The pressure
for locking was 250 bar and the pressure for un-locking was 350 bar. In this edition of locking
sleeves, a higher opening pressure would guarantee avoiding the wedge to get stuck. At the
time of the project, the Company ETP Transmission was part of the discussion concerning
how they could improve their locking sleeves in order to better support this application, which
was different in some ways compared to chucks. As a result, ETP Transmission introduced
two new locking sleeves the ETP-HYPUS 22 and ETP-HYPUS 56 that had 350/350 bar
locking/un-locking pressure. This was, in many ways, more convenient for those
implementing the system, because it was not necessary to have different hydraulic pressures
in the system. The new ETP-HYPUS 22/56 was used in the so-called Torres and Hyde
Modules. The complete hydraulic diagram for the Dynamic Modules is presented in
Appendix B. With the introduction of new locking sleeves, however, there were still some
problems with leakage. For example, if there is a drip of oil in a chuck, there is not a problem;
however, one drip of oil on an aircraft laminate material could lead to serious problems. In the
future, oil drip elimination must be guaranteed.

The Coromant Capto Interface

In positioning aircraft parts in assembly fixtures, accuracy
from end-users was required to lie within +/- 0.2 mm. That
accuracy is not at the end of the DM, but at the datum points.
Therefore, the end of the DM, described later in this
dissertation as a DM tool center point (DM TCP), must be
positioned even better than +/- 0.2 mm. All tooling used in
the aerospace industry today gets its accuracy from the
surrounding system, i.e. framework or foundations. The ART
concept uses the robot to provide the accuracy in the tooling.
This means that between the robot and the DM, no loss of
accuracy is allowed, where the loss of accuracy in the
Figure 4.10: The
74 Coromant Capto
interface
 Affordable Reconfigurable Tooling

interface between robot and DM will directly accumulate to the total error of the tooling
accuracy.

The SANDVIK Company uses a tool changing system called Coromant Capto to quickly
change cutting tools in NC-machines. The Coromant Capto system comes in different sizes,
depending on the application. The Coromant Capto C4 was used for many different purposes
in this research project, see Figure 4.10. One of these purposes was for the robot to dock onto
a DM and position with high accuracy. The repetitive accuracy of the Coromant Capto C4,
referred to in this dissertation as just Capto, was +/- 0.002 mm. This means that if the robot
manages to position a DM to +/- 0.05 mm, +/- 0.002 mm is negligible in this context. In this
research, two different interfaces were compared - the Capto and a product from SYSTEM
3R. The Capto system was chosen for its robust design, high accuracy and low price. The
Capto was also used for attaching the metrology probe as presented in Chapter 5 as a tool
changer for drilling end-effectors as was presented in Chapter 2.

The Parallel Mechanical DMs

The conceptual idea of the parallel mechanical DMs is to have a rigid structure to ensure that
tolerance requirements are maintained throughout the assembly process. A subproject
supervised by the author was made by Askbrink (2002) as part of this research to determine
which design for a parallel mechanical structure would be most suited on the force signature
defined. Askbrink evaluated several parallel structures and concluded from finite element
studies that a six-legged 3-3 Stewart platform configuration the Hexapod - was best suited
for the purpose of this research, see Figure 4.11b. The final design was further improved by
the ADFAST team and the final design and manufacturing was carried out by the Hyde
Group, who was one of the ADFAST partners.

Figure 4.11a: The Figure 4.11b: The Figure 4.11c: The

Tripod Hexapod Octapod

The Octapod (Figure 4.11c) was the first prototype designed and manufactured at Linkping
University. The Tripod was a so-called Dummy module; it was considered a waste of money
to build more than more than one of each module concept. Late in the project, the resourses
became limited, and thus the Tripod was built (Figure 4.11a) for a cost of 37. The Tripod,
which has locking devices in all six ball joints, showed a surprisingly rigid structure.

The Serial linked Mechanisms

Another DM concept are those which are serial-linked mechanisms, which essentially are
further developed Pogo-sticks that are used in CNC-controlled tooling. Pogo-sticks are 1DOF,
one axis prismatic actuators. The serial-linked DMs are flexible in 6DOF. One of the biggest
advantages of this solution is their ability to access complex structures. The Parallel mechanic
DMs are a more rigid solution, but in some cases the ability to carry loads is not as important

75
Affordable Reconfigurable Tooling

as accessing a part in a complex assembly. In this case, the serial linked DMs are more
suitable. Figure 4.12a-d shows the four different concepts of serial-linked DMs that were
designed and built for physical evaluation.

Figure 4.12a: The Figure 4.12b: The Figure 4.12c: The Figure 4.12d: The
Torres Module LiU Module Hyde Module Dummy Module

There is modularity within the Dynamic Modules, i.e. the components within the modules are
exchangeable between the modules. A serial-linked module essentially constitutes one cradle
(the first two revolute joints), which main function is to enable two rotational axes. On each
cradle is attached a Pogo-stick unit, which essentially is one prismatic joint, which main
function is to enable long reach for the module. At the end of the Pogo-stick is attached a
Ball-joint to give the last three degrees of freedom. Not in figure 4.12a, the cradle is different
from the others. The Torres Pogo-stick was to large for a normal cradle, so a carriage module
was designed to provide for higher stiffness. The Torres Module is now considered a hybrid
solution between a serial-linked and parallel mechanic structure, i.e. one which is flexible yet
rigid.

The modularity within the Dynamic Modules has an important function scale-ability. This
means that the sub-modules within the Dynamic Modules can be changed to other
dimensions. For example if one Pogo-stick does not reach far enough, it can be exchanged to
a longer version. This can be done without exchanging the hole Dynamic Module. This
thinking is employed throughout all modules and the framework as a kind of ART
philosophy. Early in the project, the objective was to have only a few building blocks that
could be up or down scaled to meet different requirements. This vision did not become fully
implemented. Instead as many different solution were designed and evaluated as possible.

The Ball-joint
The Ball-joint has 3 DOF (roll, pitch and yaw) flexibility. The idea for using the Ball-joints is
to get the last three revolute degrees of freedom of a serial-linked DM. Figure 4.13 shows a
cross section of one of the Ball-joint designs.

76
 Affordable Reconfigurable Tooling

Figure 4.13: The Ball-joint

Principally the same hydraulic system that is used for the ETP Locking sleeves are used to
lock the Ball-joint by pressing a wedge back and forth in a radial direction. As the wedge is
moving, the cup is moved axially inducing a force to the ball that is pressed against a conical
stop. The delicate design challenge is defining the relationship between the casing and the cup
to make the cup stay in the locked position as the wedge pressure drops. In addition, even
more challenging is to release the cup on demand. The Ball-joint is basically bolted at the end
of the DM. The physical evaluation is presented in Section 4.4.

The universal joint

The universal joint was designed to make the LiU Module (Figure 4.12b) and the Hyde
Module (4.12c) rotates in two degrees of freedom. Two conceptual designs were evaluated
the LiU Module had the second locking sleeve on the outside and the Hyde Module had the
second locking sleeve built in as a more compact solution.

77
Affordable Reconfigurable Tooling

4.3.3 ART Pick-ups

The interface to the aircraft parts, the Pick-ups for the ART system, is shown in Figure 4.14.

Aluminum part with

holes for interfacing
aircraft parts

Male Capto
Female Capto Chuck
Chuck

Figure 4.14: The product-specific Pick-ups

The concept of the Pick-up is based on allowing the device to be product-specific, which
means that the ART system is not setting any particular demands on the product, at least not
more than from a common sense approach. This also enables any kind of Pick-up to be
developed, i.e. anything from a hole, clamp, vacuum cup etc. The male Capto interfaces that
are used for the robot to dock onto cannot themselves be used for attaching aircraft parts
directly. To the left in Figure 4.14 is shown a pick-up with a female Capto interface, which
essentially is a small manual chuck. The female Capto interface was purchased off-the shell
from SANDVIK. One end of the interface had a female Capto interface, and the other end had
a male Capto interface. Holes in the aircraft parts were the datum points to be attached by the
ART Pick-ups. In the conceptual design, it was decided to use one and the same Pick-up for
all 8 datum holes in the aircraft parts. Figure 4.14 shows the aluminum part that is attached on
the Capto. The aluminum part has two holes in different angles to hold the aircraft parts
during assembly.

The relationship between the female Capto interface point and the holes on the aluminum
parts of the pick-ups was calibrated using a Zeiss PMC-V 850 Coordinate Measurement
Machine. It is important to realize that the aircraft parts are located by the holes on the
aluminum parts (datum points) of the pick-up (see Figure 4.14). The robot is positioning the
female Capto chuck of the Pick-up, not the datum hole itself. Therefore a lot of effort was
spent in making this calibration with extreme precision.

4.4 Physical Demonstrator

The ART was tested in a full scale demonstrator in the Production System robot laboratory at
Linkping University during spring 2004. The idea was to investigate as much as possible
given the time constraints, and to put all the building blocks together to first of all see if it
worked in practice, but especially to identify causal effects when sub-systems are put together
into one large system. A demonstrator also has the purpose to sell the conceptual idea and to
show what the project needed to accomplish.

78
 Affordable Reconfigurable Tooling

Figure 4.15: The DEMO-ART

In order to make the demonstration to visualize a realistic production scenario, a reference

airframe structure was chosen as an application for DEMO-ART, as shown in Figure 4.15.
The structure is a forward fuselage section of an Unmanned Aerial Vehicle from Saab
Aerostructures. In the design phase of the tooling, the complete forward fuselage was used as
a reference structure. However, to keep the cost down for the physical demonstration, only the
truss section was built.

The Modular Framework was put together in 8 hours, using only a hand pallet truck. Beams
were mounted one by one. What was thought to be the most challenging moment, to raise the
H-section which carries the upper section of Dynamic Modules, was in fact easy. This was
made possible using a winch attached to the top of each beam column, as illustrated in Figure
4.16a.

Figure 4.16a: The Figure 4.16b: The Modular Framework

manual winch before winching the H-section

Winching was done by four persons who synchronously operated all winches with each other.
Figure 4.15b show the H-section before winching. As the beams were in position, the screws
of the box-joints where tightened. No sliding movements were detected, although massive
weight of the upper H-section.

79
Affordable Reconfigurable Tooling

The utility of the box-joints allowed for attaching box-joints to the framework. Figure 4.17
below shows how the same box-joints, used for assembling the beams, are used as attaching
plates for the Dynamic Modules. This concept enabled the modules to be moved easily by
releasing the screws slightly and then sliding the module on the beam.

Box-joints

Figure 4.17a: A Box-joint used to attach Figure 4.17a: Box-joints

the Hyde Module to the Dynamic used for attaching parallel
Framework mechanic DMs

The beams were covered by a protective varnish that made sliding the box-joints easier.

Not all the Dynamic Modules had hydro-mechanical locking sleeves. The Octapod, the first
prototype, was locked by constantly keeping a pneumatic pressure of 700 bar. The mTorres
pogo also had a pneumatic pressure for locking, but it was passively locked, which means that
the locking sleeves only were pressurized for releasing the locking pressure. Table 4.1 shows
the locking concepts of the 8 different Dynamic Modules. Note that three DMs - DM3, DM4
and DM8 - are manual. The explanation for this is that there were not enough resources to
replicate the actively looked concepts. The Tripod modules were low-cost modules that could
be configured by the robot, but locking was done manually. The Tripods were surprisingly
stiff in the experiments. DM8 was an emulated, serial-linked module which could also be
configured in 6DOF with the robot and locked manually. DM8 was very weak in stiffness.

Table 4.1: The release and locking pressure and method

Dynamic Module Locking/Release pressure Method
(bar)
DM1: Hexapod 250/350 Hydraulic
DM2: Octapod 700 Pneumatic
DM3: Tripod dummy Friction by screws Manual
DM4: Tripod dummy Friction by screws Manual
DM5: mTorres pogo 350/350 Hydraulic and Pneumatic
DM6: LiU pogo 250/350 Hydraulic
DM7: Hyde pogo 350/350 Hydraulic
DM8: Pogo dummy Friction by screws Manual

80
 Affordable Reconfigurable Tooling

The reason for the range of locking solutions was partly because the modules were developed
in generations. As soon as one Dynamic Module was manufactured, extensive testing on that
module was performed, which in many aspects pushed development further as the next
module was to be manufactured. This also pushed the demands on locking sleeves forward.
The locking sleeves were designed to hold cutters, whereas their use in the ART application
was different. By working closely with the company ETP Transmission, soon a new locking
sleeve with opening and closing pressure of 350 bars was developed which was much more
convenient than having different pressures in the aspect of designing the hydraulic system.
The hydraulic circuit diagram is presented in Appendix B

The hydraulic system for locking and unlocking the modules consists of a 250/350bar pump,
a 400bar pump and 7 electromagnetic valves. The 250/350bar pump was used to lock and
unlock four of the five modules. Two of these modules used the same pressure (350bar) for
both locking and unlocking. This means that when the module shall be locked or unlocked, it
is only to change the direction of the hydraulic flow. For this reason, only one valve directing
the flow was needed for each of these two modules (DM5 and DM7). The other two modules
that were controlled by the 250/350bar pump used different pressures for locking (250bar)
and unlocking (350bar). Because of this, two valves were needed for each of these two
modules (DM1 and DM8).

The 400bar pump was only used to control the Octapod. The Octapod required a constant
pressure to stay locked; hence, the pump was connected directly to the module. This pump
had a pneumatic input and generated a hydraulic output that was used to lock the Octapod.
Using a back valve for the Octapod pump made it possible to turn off the pump as the
pressure was built.

The main flow from the 250/350 bar pump that came in to the Dynamic Modules was divided
into smaller hoses using a flute. This was a requirement from the locking mechanisms to use
smaller hoses instead of the bigger one coming from the main pump. This solution was new
for this kind of locking device. Normally, these locking sleeves were hand pumped. In this
case, a pump was used, which was a challenge. These kinds of locking devices have very high
pressures, but only small oil flow. High pressure gave one advantage; the air that was still in
the system was so compressed that it had no affect on the locking process.

One issue working with hydraulics was that oil leakage caused drips on the airframe structure,
which was unacceptable. Removing all of the air from the hydraulic system caused some
leakage, and tremendous efforts were spent in eliminating all possible leakage. This procedure
is done only the first time the system is installed. Some of the locking sleeves had problems
with leakage, but in future applications, ETP Transmission has promised better sealed locking
sleeves. One much appreciated feature about the system was the drip-free fast connectors used
to plug in or plug out a Dynamic Module. From the main valve come two hoses which enter
the Dynamic Module via two fast connectors; these made plug-in and plug-out of the module
convenient.

As the system was all setup, the hydraulic system was executed for days at the time to make
sure there were no leakage and no jamming effects. In some instances, the locking devices
jammed. In these cases, a hand pump was used to pump the wedge in the locking device back
to the unlocked mode. These jamming effects went away over time as the system was
trimmed.

81
Affordable Reconfigurable Tooling

4.5 ART Database

As the number of case studies preformed and the number of ART concepts developed grew, it
began to be difficult to keep track of where all the parts where located on the network. The
many CAD parts, components, beams, Dynamic Modules etc. needed a structured way of
storage to simplify finding data for the next project. In addition, starting from scratch in
designing an ART concept in a CAD system takes time. There was a need to find a method to
re-use work that had previously been done. A complex system such as CATIA does not make
it easy to move around large product files on the network without ending up with loose ends.

From this point of view, a library structure for the ART system was created which was
divided into folders depending on the modules. For example, there is a folder for Dynamic
Modules as seen in Figure 4.18. Inside that folder, products of different concepts of Dynamic
Modules can be found as well as each single part that compounds the Dynamic Module.
Moreover, the philosophy of the database is to have sample directories on each level. The
sample directories contain the work that has been done in earlier projects. To shorten
modeling time, the previous products can be combined to form new products. For example,
when designing a framework from steel beams and Box-joints, the product file is stored in the
Modular Framework sample folder, which makes it a good starting platform for future
products as well as the one it was originally designed for.

Figure 4.18: A snapshot on the ART database

The ART database is stored with a release number. In Figure 4.18, the database is called
Release 1. All applications (e.g. the ART concept) created in that release will be assured to
have an intact database without loose ends. If the database is rearranged, a new release of the
database must be created, whereas the applications in the old release cannot be moved to the
new release without extensive work. Certainly it is possible to map old applications to a new
release, but it requires remapping the CATIA product tree for that specific application.

82
 Affordable Reconfigurable Tooling

4.6 Additional Virtual Case-studies Using The ART System

The ART concept was tested on other applications except the UAV. In the ADFAST project,
there were three main reference structures that were studied: one open airframe section from
Alenia, one wing section from Airbus Spain and one body section from Saab Aerostructures
(Figure 4.15). Due to confidentiality agreements for showing the structures from partners in
the project, no pictures can be shown. Essentially what came out of that study was that ART
was possible to apply for other structures. One conclusion drawn from these studies was that a
hybrid solution between a pogo pick-up and a parallel mechanical solution had good potential,
both for reach and rigidity. This solution was later used in the DEMO-ART (see Figure 4.12a
mTorres module).

One case study was also performed outside the aerospace industry. Axelsson (2002) explained
that a car manufacturing company such as VOLVO Cars is prudent in managing at least one
new car model every year in the same production line. Axelsson postulates that flexible
production and changeovers are important factors in managing that scenario. Axelsson even
states a goal to manage starting up a new car model within 48 hours and reaching full
production within eight weeks! Based on Abelsons statements, the author of this dissertation
initiated a sub-project to investigate if ART be a technology to enable such a visionary goal.
This sub-project was presented in Wulf (2004). The case-study had a starting point from an
existing jigging solution for the VOLVO XC90 side structure. Figure 4.19a shows the
existing fixture used for holding the car structures during spot welding.

Figure 4.19a: The XC90 side Figure 4.19b: The XC90 side
panel and jig panel in ART

The sub-project concerned designing an ART system that facilitated the same building
method as the original system. Figure 4.19b is shows a concept of how the VOLVO XC90
could be fixtured in the ART system. The results from this case study showed that ART can
also be used for automotive products, but there are some conclusions that can be drawn from
this. First of all, ART is a flexible tooling system, which supports business cases where tools
must be changed perhaps once a week. Working with a high-volume product such as the
automobile makes ART somewhat superfluous. There are some interesting scenarios,
however, even for automotive industry, which could find ART an attractive solution. When
prototypes are built there are seldom fixtures available, especially where a prototype is built
before the product has been industrialised. Prototype cars are indeed a low-volume product
scenario where a reconfigurable tooling system such as ART could be applied. The same goes
for building concept vehicles, which also require tooling to fixate the parts during assembly.
A concept car is essentially a one-product scenario which would be possible to build in an
ART system. The last area is measurement control fixtures. Cars being built require a spot
check of body parts. Holding the body parts in a CMM machine requires some form of fixture
where ART can be used. If the measurements are not more demanding than 0.05 mm, the
robot with a metrology system brought about in the research presented in this dissertation

83
Affordable Reconfigurable Tooling

could be used. Thus, Axelssons vision to build the product in 48 hours using the ART system
would perhaps be possible. In one of the case studies that were presented in the subproject,
supervised by the author (Wulf, 2004), the starting point was to have no previous ART system
to modify. Wulf had learned to work with the CAD/CAM system for 10 weeks. The ART
concept that is shown in Figure 4.19b took several weeks for Wulf to design. But Wulf
concludes: given the experience after this project the same concept could have been designed
in one working week. Note that this time is not including the modelling of the car parts. It
must be noted that at the time the sub-project was executed, the ART methodology was under
development. In addition, Wulf had no previous experience working with CATIA and
DELMIA, which were the applications used for design, simulation and programming in the
sub-project.

4.7 Economical Aspects of the ART System

In Paper II in the Appended Publications, (Herbertsson et al., 2002), the economical aspects
of ART presented. The conclusion from that paper was obvious: if a reconfigurable system,
such as ART, technically worked, and given the right scenarios, it would be likely that the
ART system would provide an economic pay-off. The economical study was based on
comparing conventional tooling, as a reference case, with an estimation of ART, in order to
see where the break-even point occurred. The estimation of investment costs for ART were
made on a pessimistic basis, i.e., costs were overestimated rather than underestimated.

The study concluded that ART has a break-even point at 3 conventional tools, i.e., it will pay
off if 3 conventional tools can be replaced by 1 ART tool. In practice, however, due to the net
present cost (taking interest rates in accordance), it is more likely that ART must replace 4-5
conventional tools to reach this break-even point over a 15-year time period. In the scenario
of investing in tooling for a total volume between 200 and 400 units over a 15 year period, it
was assumed that 2 ART systems were needed for years 1 and 3, compared to 8 conventional
tools over the same period. Taking the net present value with 5% interest, conventional
tooling would cost 1,345 (money units) compared to 1,190 (money units) for ART. Higher
interest rates would make conventional tooling seem more profitable. This is because,
economically, it is worse to have to invest more money directly instead of spreading the
investment over a longer period. In this aspect, it is important to have in mind that investing
in an ART tool is not likely to be used for only one project over a 15 year period.

Another important issue when making an industrial investment calculation is to look at other
qualitative advantages that may arise due to the investment. In the scenario described above, a
number of likely positive aspects were not taken into the quantitative analysis, but would
probably have significant positive effects for ART in reality. These included:

Better possibilities to level out variations in the production rate due to increased
flexibility
Efficient utility of floor space
Easier and cheaper ways to make changes in the airframe design later in the product
life cycle
Possibility to rapidly prepare and reconfigure for new assemblies since a lot of the
work can be performed on a computer, offline
Shorter lead time in tool design and in building the tool
Possibility to use one system for many different projects

84
 Affordable Reconfigurable Tooling

This section concludes the economical discussion of ART in this Chapter. Before moving on
with more enabling technologies in this dissertation, a discussion is made to point out how
ART relates to similar research attempts.

4.8 Similar Research Attempts to Robot-manipulated Tooling

In this Chapter, a technical solution using a robot to reconfigure a tool has been presented,
and the author has tried to explain both advantages and disadvantages with the approach. This
section of the chapter has the purpose of further analyzing the approach by comparing ART
with other solutions that have similar characteristics.

Although the author has not found any application in the literature or industry that has done
quite the same, the fundamental idea of using a robot manipulator to reconfigure a tool has
been done before. One intriguing approach on the matter has been done at McDonnell
Douglas in Long Beach California, which merged with Boeing in 1997. McDonnell Douglas
developed an automated flexible assembly cell in order to reduce tooling cost. The installation
was an automated cell of subassemblies, and the system was able to replace up to 250
conventional tools (Goodman, Dec. 1990). One single flexible tool could serve a family of
assemblies and provide automation inspection. This was accomplished by using a gantry
robot to reconfigure the tool between the subassemblies and automatically load detail parts
into the jig. After the tool was configured and the parts were inserted, the system
automatically drilled and fastened the parts into subassemblies. In addition, prior to moving
the assemblies out of the system, the parts where verified through a vision system (Goodman,
Sept. 1990). One of the issues with the system was that the ambition to fully automate the
process resulted in too much complexity and interruption due to frequent technical problems.
This concerned especially the reconfigurable tool and its clamping force to hold the parts in
place (Alam, 2005). Moreover, most of these kinds of subassemblies are fully integrated parts
today. Mr. Alam, the project manager of the implementation, further explained that inspecting
every detail was a bottleneck in the process and raised the question whether it was an efficient
process flow or not. The concept generated by Boeing was impressive and would probably
have worked if all technical problems would have been solved and if a powerful, fast
computer had been included, as well as a production hardened system with an efficient
process flow (Alam, 2005). The fundamental approach from Mr. Alam and Mr. Goodman has
many aspects, and has made a solid contribution to this research. The two major differences
between the project by Boeing and this research are the selection of robot and level of
automation. Boeing used a large gantry robot, whereas the author used an IRB 4400.

Another research group that used a robot to configure/reconfigure a fixture was carried out at
the Massachusetts Institute of Technology (M.I.T.) in the late 80s (Youcef-Toumi et al., 1988;
Youcel-Toumi and Buitrago, 1988; Asada and Fields, 1985; Gordon and Seering, 1988). The
M.I.T. project involved the automation of sheet metal drilling using a lightweight robot.
Robotized drilling of sheet metal is considered difficult because of the complex geometries
involved. The M.I.T. project succeeded in using a method called Robot-operated Adaptable
Fixture in order to change the configuration of a fixture bed by manipulating vertical fixturing
units, quire similar to the bed-of-nail concept, using Pogo sticks, presented in Section 4.2.3.
By fixing the sheet-metal parts into the Pogo bed, the team managed to automate the drilling
prior to assembly. The research at M.I.T. is similar to the ART concept in the way that pick-
ups are configured by a lightweight robot manipulator, but first of all the pick-ups are one-
axis in this research and the complexity is one dimensional access to the airframe parts only,
whereas the Dynamic Modules in ART are flexible in 6DOF. The ART concept is also
modular and can take it all the way to a three dimensional access scenario.

85
Affordable Reconfigurable Tooling

The last research attempt to robot-manipulated tooling, such as the ART concept, was also
made by Boeing but more recently. The two reports that the author has found on the topic
were presented by Munk (2002) and Stone (2004). Their approach to reconfigurable tooling
was based on using a 5 axis, CNC-directed traveling-column post mill. The machine was used
to reconfigure a fixture in three degrees of freedom. The fixture consisted of 56 passive Pogos
located on 25 towers, with two per tower on each of 19 towers and three Pogos on each of six
towers. In this configuration, the NC machine docks with each of the Pogos, and when the
controller releases the brakes on the towers and Pogos, the machine moves the Pogo in three
axes to the desired location. The task for the team designing the system was to create a fixture
that would be low cost, reliable and yet flexible enough to handle the various existing and
future spar designs. The NC machine was also capable of drilling, rapping, machining,
trimming and probing. The system had undergone extensive testing and has assembled several
full-sized spares to verify its performance. This approach on robot-manipulated tooling from
Boeing has the same fundamental approach, but Boeing is using a NC-machine to configure
the Dynamic Modules instead of an industrial robot. In addition, the Dynamic Modules in the
Boeing project were flexible in 3DOF, whereas in the ART concept 6DOF. One advantage
compared to the ART concept is that the work volume is much better covered in the Boeing
project. Comparing that perspective, the robot could be put on a track to move into a larger
work volume. The Boeing project, using a rigid NC-machine, does not need to use a
metrology system such as the ART Concept. The next Chapter in this dissertation will explain
how the metrology system was integrated to the robot controller.

86
PART III: METROLOGY-INTEGRATED ROBOTICS FOR
AFFORDABLE AUTOMATION

PART III
METROLOGY-INTEGRATED ROBOTICS
FOR AFFORDABLE AUTOMATION

In the previous Part II, industrial robots were utilized for performing drilling
automation and reconfiguration of a flexible tooling concept. Part III presents
how a metrology system, normally used in aerospace manufacturing for fixture
calibration, is used to provide the high accuracy demands put out by the airframe
assembly process. Chapter 5 introduces a different potential metrology system
followed by the in depth presentation of the metrology system used for this
research. Chapter 6 presents how this research has integrated the metrology
system with the robot controller in a feedback loop to guide the robot to high
accuracy. Chapter 7 is describing the mathematical derivation of calibration
procedure as part of the integration software developed in this research.

87
88
 Metrology for Maintained Accuracy

5 Metrology for Maintained Accuracy

From what was concluded in Chapter 2; Industrial robots are just about to achieve the
accuracy required for airframe assembly, but furthermost, that accuracy is for static force
prerequisites. The reconfigurable tooling presented in Chapter 4 has a dynamic force
scenario when the robot is manipulating the Dynamic Modules. Chapter 3 concluded that a
pressure foot is necessary although the forces during Orbital drilling are low, where the pre-
pressuring force will cause the robot to deviate from its perpendicular orientation to the
surface, depending on the actual configuration of the robot. This Chapter is introducing
existing metrology systems that have the potential for being incorporated with the robot
controller to provide high accuracy.

Encyclopedia of Britannica Online (2005) defines metrology as:

Metrology is the science of measurement. From three fundamental quantities,

length, mass, and time, all other mechanical quantitiese.g., area, volume,
acceleration, and powercan be derived. A comprehensive system of practical
measurement should include at least three other bases, taking in the measurement
of electromagnetic quantities, of temperature, and of intensity of radiatione.g.,
light.

What makes robots reach high accuracy using a metrology system is neither repetitive nor
absolute accuracy itself. What is important is the resolution of the robot. Resolution on the
other hand, often goes hand in hand with high repetitive accuracy. Let us look at the robot
used for the experiments - an IRB4400 from ABB. The repetitive accuracy in that robot is
0.07-0.1 mm (ABB IRB4400, 2004). To reach that accuracy in the full work volume, the
resolution of the robot is much higher. Tests by the author have proven the robot to be able to
move with a resolution of 5 m. When fine adjustments have been carried out, the robot
controller is simply given a new absolute position in the base coordinate system.

Many different ways have been tried to compensate for dislocations in robots today (Gooch,
1998; Beyer, 1999), i.e. where an external measuring loop is used to improve the final
position of the robot TCP (Tool Center Point). These methods basically move the robot to a
position in space, and when the position is reached, the measurement system can verify and
correct the position. Gooch (1998) stated that metrology is an enabling technology and
applications are being extended from inspection to control of the manufacturing process itself.

In aircraft assembly, metrology is used to control the relative positioning of different

components during the assembly task. Earlier, positioning of components was done using
physical gauges, but now is more commonly achieved using measurement instruments that
measure the distance and angles to targets placed on the aircraft components. These values are
then compared against the desired positions (and tolerances) defined in the CAD data. Such
instruments include theodolites, photogrammetry systems and laser trackers.

This section of the report will give the reader an introduction to different kinds of metrology
systems that are used and has relevance to the one metrology system used in this research.

89
Metrology for Maintained Accuracy

5.1 Metrology Possible for Robot-integration

5.1.1 Laser Trackers
Laser trackers provide a relatively fast, precise and intuitive method of measuring large
objects in industrial environments.

Functional description
A laser tracker measures in three degrees of freedom (3DOF) the three-dimensional location
of a mobile target with an accuracy of a few micrometers, over a range of tens of meters. A
laser tracker is fast and accurate for measuring large targets that can be moved anywhere in
the measurement volume. A laser tracker is an alternative to Coordinate Measuring Machines
(CMM).

A laser tracker uses laser interferometry to measure the distance. The light source is split into
two beams by a semi transparent mirror. One beam is used as a reference and is reflected on a
detector by a reference mirror. The second beam goes out of the tracker unit and is reflected
back to the detector by the moving target mirror, as seen in Figure 5.1. The light interference
between the two beams is measured, making the unit a laser interferometer. The number of
fringes is counted as the external path length changes. Note that this method is restricted to
linear measurements, and hence 1 DOF.

Reference Mirror

Beam Splitter
Laser Source

Moving
Mirror

Interference fringe
Detector patterns

Figure 5.1: Principles of a Michelson interferometer set-up

The other two degrees of freedom, the azimuth and elevation, is achieved by using beam-
steering mirrors to direct the beam in a wide range of directions. The moving target mirror is
called the prism retro-reflective target, or in laymans terms, a prism. To have the beam
pointed at the prism, a feedback loop to control the moving mirrors in the tracker unit is used.
When the laser beam hits the prism target off-center, it is reflected back parallel to the
incident beam, but displaced. A two dimensional sensor measures the displacement, allowing
the laser tracker to adjust the beam-steering mirror and return the beam to its desired coaxial
state. This can be seen in Figure 5.2.

90
 Metrology for Maintained Accuracy

Laser beam Prism

Beam displaced

Coaxial state

Figure 5.2: Tracking mechanism using the prism

When the beam hits the center of the target, it returns without displacement indicating the
beam has hit the correct location. One of the advantages with laser tracking is that it can
measure with high frequency, although the target is moving at several meters per second.

Given in polar co-ordinates, the distance, elevation

and azimuth (see figure 5.3), are transformed into
Cartesian co-ordinates, which are the common
representation for calibration and robot applications.
To measure Cartesian locations in the measurement
volume, an operator walks around and positions the
prism in different locations. The measurement
process starts by taking the prism from the
instrument datum point called bird bath. Since a
laser interferometer only measures relative distance,
a well known start position is required. If the beam Figure 5.3: The distance,
is broken by mistake the operator has to put the elevation and azimuth
prism in the bird bath again before measurement
continues.

The typical applications for laser trackers are:

robot calibration
shipbuilding
aircraft manufacturing
fixture build and calibration
verification of the design of manufactured structure
reverse engineering
inspection and alignment

In the process of using a laser tracker, a large amount of information is gathered. CAD-
models that correspond to the nominal data can be compared with the real physical object
using the laser tracker. The laser tracker can gather information of a surface by measuring in
continuous mode. Normally the stationary mode is used to get one measured point from the
tracker unit. If continuous mode is active, a stream of data is achieved. A laser tracker has
software packages which can be used to create mathematical geometries such as lines, curves,

91
Metrology for Maintained Accuracy

arcs, circles etc. Advanced statistical analysis is also available in these programs. A laser
tracker is also well suited for measuring deformation and movement of the components. From
the tracking abilities the tracker can continuously measure a deformed or moving target.

There exist different kinds of prisms. The most commonly used prisms are the open air corner
cubes that were presented in Figure 5.2. The open air corner cube has an acceptance angle of
+/- 20. Another innovation in reflector technology is to fill the space inside the prism with
glass. This gives the beam a refracting effect when entering the glass as seen in Figure 5.4a.
This method increases the acceptance angle for the prism. The effective target points are no
longer the fixed intersection point of the reflecting surface but are instead a shifted virtual
point, whose shift is dependent on the angle of the laser beam to the prism.

Virtual Refracting
point effect

Lateral
deviation

Center of tooling ball

Figure 5.4a: A prism Figure 5.4b: A prism

with additional glass with additional glass

Using glass prisms however, does not come without problems since accuracy dependencies
can be observed in respect to the entry angle of the laser beam. This matter is thoroughly
investigated by Markendorf (1998) and Kyle et al, (1997). Essentially this means that the
more angles contained on the prism the less accuracy there is, as can be seen in Figure 5.4b.

Another reflector used is the Cats eye reflector with an acceptance angle of +/- 60. The
Cats eye reflector was the first retro-reflector used when laser trackers where developed. The
concept of the Cats eye is to have two glass hemispheres glued together. The advantage with
this solution is the high acceptance angle they can possibly have. The disadvantage is the
relatively large size and weight of the device.

Laser tracking technology has the advantage of rapidly measuring a target in space. The
obvious shortcoming is that the laser beam cannot be broken.

5.1.2 Photogrametry Systems

Photogrammetry is the traditional name for 3D Image Metrology, which is a technique to
measure three-dimensional coordinates, using Images and the principle of triangulation
(Johnson and Walker, 1998). Photogrammetry systems measure to points by optical
triangulation seen from more than one camera view in the volume that ranges in general from
0.1m3 to 30m3.

In Figure 5.5, each camera is taking pictures at the same time stamp. In normal image
technology, a process of analyzing edges or any features are used. In photogrammetry, and for
high accuracy measurements, an algorithm is used to find projected spots of lights either self
92
 Metrology for Maintained Accuracy

illuminating targets, such as light emitting diodes, or a retro-reflective target. Each of these
targets is identified in the respective camera picture. Another algorithm is combining these
patterns to 3-dimensional coordinates, which is called triangulation. To achieve a scale on the
object, a reference object has to be measured. This reference object can in addition give the
datum definition of the measurement volume, hence the origin (0,0,0). This origin can be used
as a zero reference for a common global coordinate system.

Camera 1 Camera 2
x1,y1 x1,y1
x2,y2 x2,y2
x3,y3 x3,y3
x4,y4 x4,y4
x5,y5 x5,y5
x6,y6 x6,y6
x7,y7 x7,y7
x8,y8 x8,y8
x1,y1,z1
x2,y2,z2
x3,y3,z3
Combining x4,y4,z4
patterns to a 3D- x5,y5,z5
picture x6,y6,z6
x7,y7,z7
x8,y8,z8

Figure 5.5: The Concept of Photogrammetry

Typical accuracy is in the range of 10 to 100 micrometers. The cameras can either be in a
fixed configuration, where the measurement system can be treated as black box requiring little
operator expertise, or a single roaming camera can be used to take multiple images from a
variety of viewpoints. Although each camera can take many pictures per second, it is not
feasible to say that more than 25 3D-positions can be given per second.

One of the most advantageous aspects of using photogrammetric metrology is the capability
to provide multiple measurements at one instant in time and for this process to be repeatable
as often as 100 times a second. This means, the 3D deformation of an object could be
measured in real-time. With photogrammetry is possible to have redundancy of references on
the object being measured, should another object shade the cameras. Measuring several
references simultaneously allows the measurement of multiple objects relative to each other.

Krypton
One suitable system for robot-integrated metrology is the Krypton system from the Belgium
company Krypton. Their latest addition to metrology is the K610 seen in Figure 5.6. The
K610 sensor is a pre-calibrated and a certified camera measurement system. This means that
in the field, calibration of the camera system is not necessary. The camera is composed of a
ridged structure with three linear Charged Coupled Devices (CCD) cameras. These cameras
have the advantage of high resolution (2000 pixels). A pixel interpolation algorithm for sub
pixel resolution enables the system to obtain a resolution of 1 part in 200.000. Dynamic
measurements can be taken up to 1000 Hz for 3 LEDs.

93
Metrology for Maintained Accuracy

With this system, Infrared-LED 3D

positions are measured. With 3 or more
of these LEDs (markers), a 6 DOF
measurement can be obtained.
Measurements can be taken up to 6000
mm away from the camera unit.

The Robot L- Probe (Figure 5.7) is

used to make position measurements of
industrial robots with the K-series
system. Figure 5.6: An overview of
the Krypton system
Using 3 LEDs and one multiplexer, the
camera system can track the position of a pre-calibrated
reference point (for example the TCPF of a robot). In
combination with the K-series 6D system, this probe
makes it possible to perform accurate 3 dimensional
measurements in a very large volume. Only one flexible
cable connects this device to the controller, which makes
it very easy to operate. Rotating the robot around its TCP
allows the calibration of a new reference frame. The
Rocal software takes measurement points automatically.
The Robot L-Probe has a large base plate that makes it Figure 5.7: The L-probe
easy to attach to a robot end-effector.

Figure 5.8 represents the volume in which the K610 camera can measure objects. Due to its
specific construction with line-scan CCD sensors and cylindrical lenses, 3 different zones can
be described. Each zone has its well-defined volume and accuracy
specifications.

Cameras

Figure 5.8: The accuracy zones for the Krypton K610

The accuracy of the K610 system depends on the measured length and the distance between
the object and the camera. The measurement range is divided into 3 accuracy volumes as
indicated in the previous graphs. It is good practice to carefully choose the position of the
camera relative to the object. This always enables the user to measure with the highest
accuracy. Table 5.1 shows how the accuracy is calculated for the K610 system:

94
 Metrology for Maintained Accuracy

Table 5.1: The accuracy for the Krypton K610

Distance to the camera Zone I : Zone II : Zone III :

1,5 to 3 m 3 to 5 m 5 to 6 m

3D measurement
uncertainty 0,060 + 0,010.L 0,070 + 0,025.L 0,140 + 0,025.L
(U95)

For example, calculating the measured object with the length 1 meter in Zone I, the accuracy
can be calculated as: 0,060 + 0,010 x 1 = 0,07 mm.

The indicated measurement uncertainty is expressed for a confidence level of 95%, according
to the ISO 10360 II, VDI 2617 and ANSI / ASME B89.1.12M standards for acceptance of
CMMs.

5.1.3 Indoor Positioning System

A new, large-scale metrology system is the Infrared Positioning System (IPS). The IPS is also
known as Indoor GPS (Kang et al., 2004), which however, is not an appropriate name. IPS is
analogous to GPS, where GPS uses a matrix of satellites orbiting our planet emitting signals
to GPS receivers. IPS uses infrared light and lasers emitting signals in a relatively limited
volume; hence IPS is not a global positioning system. The IPS triangulates the positions of
receivers in a local volume, e.g., a room, by using transmitters that emit laser light and
infrared light. Receivers located inside the measurement volume detect the signals with
photodiode detectors. These photo detectors pick up the signals and compute angle and
positions based on the timing of the arriving light pulses. One, existing IPS solution is the
Constellation3Di from the company ArcSecond (ArcSecond, 2004). Constellation3D is built
on using two rotating lasers transmitting laser planes with 90 degrees angle in between, see
Figure 5.9a.

As the head of the transmitters rotates, the receivers measure the time between the initial and
subsequent pulse as each laser plane is detected. The final value provides the vertical angle
between transmitter and receiver. The rotating head on each transmitter rotates around 40 Hz.
By slightly varying the rotating speed in each transmitter the receiver can determine between
the transmitters being detected. Two pulses of infrared light are emitted twice every rotation
and with reference to the rotating lasers provide the azimuth angle to the receiver as seen in
Figure 5.9b. Each transmitter provides a one-way single path to the receivers. Two or more
transmitters are used to triangulate the coordinates of the receivers.

95
Metrology for Maintained Accuracy

30
Az

El

90

Figure 5.9b: Azimuth and

Figure 5.9a: The two elevation are measured to
rotating laser planes the receivers

Hypothetically, one receiver could detect two or more transmitters simultaneously, and
therefore confuse which transmitter is which. This is, however, a rare situation. Also if the
receivers are positioned on the same level as the transmitter, then there would be a very small
time difference between the laser planes and the error would increase.

DeLand (2004) presented the use of ARC Second at Boeing in their aligning and assembling
of large structures. Boeing has a patent application in this matter. Hawker de Havilland (HdH)
automation research in Melbourne Australia (Newberry, 2005) is using IPS for positioning
robots. Experience at the HdH has concluded in order to obtain an accuracy of 0.2 mm for a
work-volume of 5-20 x 5-20 meters, 3-4 transmitters have to be positioned throughout the
work volume. Practically, one transmitter can gather one-way position information, hence the
relative azimuth and elevation from the transmitter to the receiver. Together with a second
transmitter the position of the receivers are achieved in the base coordinate system. Two more
transmitters maximize the accuracy by lowering the triangulation errors. Kang et al. (2004)
concluded that IPS provides 4-8 parts-per-million (ppm) accuracy under controlled
conditions. This is equivalent to 4-8 m/m. Experience at Hawker de Havilland automation
research in Melbourne shows that the minimum distance between transmitter and receiver is 5
meters (15 feet) to avoid the infrared pulse overloading the receivers and resulting in
nonsensical measurements.

Experience at HdH has shown that the system can be problematic if the transmitters are not
carefully located in the work volume. However, great flexibility and large work envelopes
show the technology will be promising (Newberry, 2005). Another advantage is that the
number of detectors is independent in contrast to laser tracking; hence all detectors can be
located simultaneously. The disadvantage however is that the accuracy using IPS today is
lower than a laser trackers and the setup time is long in comparison. However, the IPS
becomes far more effective if positioned in a set location where it can be correctly adjusted
and left in position. In addition, calibration may be automated when sufficient equipment is
available, which further improves the use of the system. Finally, one of the greatest
advantages using IPS is the fact that the transmitters do no need to know where, or how many
receivers are in the measurement space. Consequently, no restarts etc. are necessary if a
receiver becomes shaded by another transmitter, wherein the measurement space is effectively
saturated with laser/infrared pulses.
96
 Metrology for Maintained Accuracy

5.2 Metrology Systems Used In This Research

5.2.1 The LTD800
Normally Laser Trackers are based on measurements of a single reflector. Where laser
interferometers, in general, must track the reflector relatively from a fine calibrated start
position (birdbath), if the beam is broken or another prism is to be measured, the tracker
requires re-calibrating the prism in the bird-bath. The LTD800 however, uses an ADM
(Absolute Distance Meter) that complements the interferometric system (see Figure 5.10).
The ADMs measurement technique is based on a systematic modulation of a light beam, like
those found in other Electro-optical Distance Meters (Kyle et al., 1997). The ADM enables
the tracker to locate individual target reflectors without the need to measure from a calibrated
origin position. The ADM can independently measure the absolute distance to any unknown
prism location and simplifies re-initializing the interferometer. Where the ADM is only used
in initializing the interferometer, it cannot be used for tracking by itself. The ADM is
initialized by a spiral-search. The spiral-search rotates
the laser beam over an area close to a prism until it detects
and locks to the center of the prism. This takes
approximately 5 seconds for each target to be detected.
Even if the exact position is known, a spiral search with
radius close to zero is carried out every time the ADM is
used. The ADM is beneficial for applications, such as
periodic inspection, where the same points on an object are ADM
to be measured. The ADM then enables a quick move
between points during measurement. A similar application
is using the tracker for building and assembling. In this
case, an array of targets can be set up on the first part as a
reference. Other parts can then be measured quickly and
iteratively relative to this reference. Automated systems Figure 5.10: The ADM
are another area where the ADM is useful. The ADM can
be used with Numerical Controlled (NC) machines or
robots for calibration, or on-line to improve their accuracy. The most evident advantage using
the ADM is when working in congested areas where obstacles break the line-of-site between
the tracker and the target. Without the ADM, the operator would need to place the prism in
the bird-bath every time the line is broken. One disadvantage with the ADM is that accuracy
is dropped slightly when it is used. Another restriction when using the ADM is the minimum
measuring distance on 2.0 m. The latter restriction is caused by the concept of separating the
interferometer beam, with a wavelength of 632.8 nm, from the ADM light. This is
accomplished by having an infrared laser diode with a wavelength of 780 nm as the ADMs
carrier beam. The modulation is done by applying an alternating current to an optical crystal
with a frequency set between 750 MHz and 920 MHz. A minimum bandwidth of 150 MHz is
necessary to achieve the shortest measuring distance of 2.6 m for the ADM, which reduces to
2.0 m when integrated into the tracker unit (Kyle et al., 1997). The LTD800 measures with an
accuracy of 10 parts per million (ppm, i.e. 10 m/m) for static targets, and 20 ppm for
moving targets. Distans resolution is 1.26 m.

97
Metrology for Maintained Accuracy

5.2.2 The T-Cam and T-Probe

In the ADFAST project the company Leica released their new addition to trackers to provide
6DOF measuring. Lasers, as presented in section 5.1.1, provide only a point in space. To
measure orientation in 3DOF, a camera was used, called the T-Cam see Figure 5.11. The T-
Cam, attached on top of the tracker unit, measures the orientation of a target, see Figure 5.11.

Orientation in space
from LED images

Position in space by
reflector tracking

Tracker with T-Cam T-Probe

Figure 5.11: Probing with the LTD800

The T-Cam uses the T-Probe as a hand-held device to probe points similar to a Coordinate
Measurement Machine (CMM). The T-Probe comprises of a glass prism for laser tracking and
10 LEDs. The LEDs is flashing infrared light in 100 Hz, which enables the camera filter out
background according to conventional photogrammetry. Angular accuracy of the T-Cam is
0.02 degrees. Note that this accuracy is not dependent on the distance away from the tracker
unit. This is because a zoom objective is used. The zoom objective continiously maximizes
the resolution from the reflector targets in the image picture. When the laser tracker has found
the prism through a spiral search and the ADM, the tracker calculates the distance to the
probe using the interferometer. From that point, the tracker communicates with the T-Cam to
zoom accordingly to that distance. At this point the T-Cam gets full resolution in the camera
picture and the accuracy can be kept constant. The T-Cam need not continuously have all 10
LEDs in the lign-in-sight, which would be cumbersome. However, accuracy will drop when
less LEDs are visible in the camera picture.

5.2.3 The 6DOF Metrology Reflector

To measure robot positions in 6 DOF a reflector was developed. The 6DOF Metrology
Reflector, in this report called 6D-Reflector, is a further development of the T-Probe that is a
commercial product from Leica Geosystems AG. The 6D-Reflector is always initiated with
the ADM, since there is no birdbath available for the 6D-Reflector. The 6D-Reflector
comprises of an aluminum housing that is attached 10 LEDs and a prism see Figure 5.12. The
LEDs enables the T-Cam to measure the position and orientation of the 6D-Reflector similar
to the T-Probe.

98
 Metrology for Maintained Accuracy

2 out of the 10
Female Capto LEDs

Male Capto

Figure 5.12: The 6D-Reflector in two angles to

show the female and male Capto

The LEDs are positioned at different depths for the camera to measure orientation. The LEDs
are flashing in infrared light and the camera is zooming in the 6D-Reflector so that the LEDs
are taking up the full picture. The laser beam is reflected back to the tracker unit from a glass
prism retro-reflector. The glass enables the beam to be +/- 50 degrees away from the prism
axis before contact is lost. The 6D-Reflector housing has two ergonomically designed handles
when moving the 6D-Reflector manually.

The 6D-Reflector was brought about to serve several purposes. The first and most important
purpose was to attach the reflector on the robot, to measure the positions of the robot, see
Figure 5.13a. The idea was to have the LTD800 track the robot while it was moving and
compensate for deviations, by having the 6D-Reflector next to the chuck. A secondary
purpose, to calibrate the chuck was also implemented. Since the Capto interface had both a
male and female interface, the male interface could be used for attaching the 6D-Reflector to
the chuck, see Figure 5.13b. The third purpose using the 6D-Reflector was to probe the
Dynamic Modules, see Figure 5.13c. In this case the Reflector was attached on the Modules
and probing their locations.

Laser beam
Laser beam

Laser beam

Figure 5.13a: Figure 5.13b: Figure 5.13c:

Continuous control Chuck calibration Probing DMs

The robot chuck constituted of an automatic, hydro-mechanical Coromant Capto chuck

system. This chuck had a female configuration and the Dynamic Modules has a male
configuration. To use the Reflector for both these cases, the Reflector had to have both a male
and a female chuck system. It was considered too expensive and too bulky to have the same
chuck solution as was used in the robot. Therefore an adaptor component was chosen as
chuck solution. This adaptor, see Figure 5.14, has both a male C4 Capto interface, and a
female end.

99
Metrology for Maintained Accuracy

Figure 5.14: A Coromant

Capto C4-Adaptor

The Capto system has an repetitive accuracy of +/- 2 m and are designed to handle loads up
to 17 kN. The adaptor was installed in the center of the Reflector unit. The reflector also was
designed with two handles for ergonomic purposes, such as probing and attaching it on the
robot.

5.3 Similar Concepts to This Research

Guiding a robot controller that has been introduced in this Chapter is not new in itself.

One system using Tricept robot in the same context as this dissertation is the TI2 system see
Figure 5.14. The TI2 was developed by Imetric SA in cooperation with the aerospace partners
Boeing, BAe Systems, Rolls Royce and Bombardier. The system was certified by Boeing to
control a Tricept robot in 1999 (CUMULI, 2000). It was stated that this was the first attempt
in the world where a highly accurate NC machine is controlled in 6 DOF for machining
operations by a 3D Image Metrology System.

100
 Metrology for Maintained Accuracy

Cameras

Machining spindle

Figure 5.15: The TI2 Concept using metrology and Tricept robots

The solution, called TI2, got its name from the products used in the initial product
development, i.e. a Tricept robot from NEOS Robotics, an Imetric 3D Image Metrology
system from Imetric and the IGRIP simulation software from Deneb (Fayad et al., 2002).
Whinnem and Nystrom (2000) further presents the TI2 concept being concepted by a
consortium of small companies who teamed up with Boeing Phantom Works to develop an
intelligent system capable of meeting the needs for aerospace assembly, essentially using the
Tricept robot online with the 3D Imaging system to improve, compared to conventional robot
system:

Stiffness
Repeatability
Positional accuracy
Portability
Operation to CAD data

The metrology system in TI2 enables in-process inspection, such as hole positions, trim paths
and other manufacturing operations. This information however, is not directly feed back to
update the machine path but to provide data for an inspection report and statistical process
control.

The difference between the TI2 approach and the approach taken in this research is mainly
that the TI2 is dedicated to Tricept robots, whereas the research presented in this research uses
an anthropomorphic robot. Furthermore, the TI2 system was used for machining, whereas this
dissertation presents robot-manipulated tooling.

Another project actually using an ABB robot and a Leica Tracker was taken on by Hawker de
Havilland (HdH). HdH uses the tracker to fine adjust a drilling machine to high accuracy. The
concept at HdH is based on moving with the robot roughly to a location for hole drilling.
Inside the drilling end-effector are two extra degrees of freedom that is used to fine position

101
Metrology for Maintained Accuracy

the drill bit by the tracker. The fine tuning is achieved from two electric actuators with a
relatively fine resolution. The project is going into production during 2005 (Crothers et al.,
2004 and Crothers, 2005). The major difference between the HdH concept and the LiU
concept is that the tracker guidance at HdH is for hole drilling only. In addition at HdH, the
tracker guides the drilling machine and not the robot itself which is done in the LiU concept.

102
 System Integration

6 System Integration

Chapter 6 presents how the integration is implemented and how the robot and metrology
system are interlaced within a network

6.1 Embedded System Control

The general definition of an embedded system is, the combination of computer hardware
and software, and perhaps additional mechanical or other parts, designed to perform a
dedicated function. The research presented in this dissertation integrates a metrology system
(the LTD800) and a robot manipulator (ABB IRB 4400) together with their respective
controllers. This section presents both the WebWare system for robot integration and the
emScon system for the tracker.

6.1.1 WebWare For Robot Integration

The robot was embedded in a windows based Ethernet environment using ActiveX within
Visual Basic. ActiveX is not in itself a programming language but rather a set of rules
governing how applications should share information. Programmers can develop ActiveX
controls in a variety of languages including C, C++, Visual Basic or Java. An ActiveX control
is similar to a Java applet. However, unlike Java applets, ActiveX controls have full access to
the Windows operative system. This gives them much more power than Java applets. Unlike
ActiveX, Java applets are independent of the operating system platform and are not limited to
the Windows environment.

WebWare is a software platform, enabling local and remote monitoring and control of plant-
wide production data, statistics and diagnostic information using standard Internet browsers.
This enables users to connect to their local area network, dial in from a remote location or log
into the Internet and analyze live and historical production data. Robots are connected to the
WebWare server and users can constantly monitor their operations and collect data and
logging errors. The WebWare package enables users to customize their own website and
tailor any system for the collection of process-specific data. This research does not use actual
WebWare software package, but instead uses WebWare Software Development Kit -
WebWare SDK.

WebWare SDK is based on the ABB InterLink communication module. The ABB Interlink
uses the Robot Application Protocol (RAP). This protocol is installed in the option pack
FactoryWare and can be purchased as an addition to the ABB S4+ robot controller. One
particularly useful feature within RAP is the Spontaneous Messaging. Spontaneous
Messaging is updating variables only when an event occurs; hence all applications that are
subscribing on the controller event become updated only after an event. This method allows
RAP to reduce the amount of communication overhead and thus no polling for data is
required. As a metaphor, one does not have to lift the telephone receiver to check if someone
is calling. Spontaneous Messages are issued as a result from the following type of events:

Logged system events and errors, such as a change between manual to auto mode
Digital input channels changes

103
System Integration

The RAPID SCWrite instruction may transmit a RAPID program variable to the
remote PC

WebWare also includes InterLink, which is an embedded technology that manages

communication with all connected ABB robot controllers, see Figure 6.1.

S4 OPC Server WebWare WebWare

SDK Server

ABB Interlink
Module

Figure 6.1: The ABB InterLink Module

It is the InterLink Module that creates, maintains, and tracks the status of the communication
link to the robot(s). The InterLink Module notifies the application of any change in the
communication link via events and properties.

The WebWare SDK is used by incorporating ActiveX controls within an application. The
operation of each control is configured via the controllers properties. In the implementation
of this research using WebWare SDK, four robot specific ActiveX controls have been used:

Helper control: is the primary interface engine for The WebWare SDK. The Helper
control provides methods, properties, and events to expose the entire S4
communication interface.

Button control: provides a way to view and modify a digital signal, and in most cases,
can be used without adding code to the application. The button action can be
configured to turn a digital signal on, turn a signal off, toggle a signal, or pulse a
signal.

Pilot Light control: tracks the state of a specific digital signal without adding any
code. Essentially, this control is similar to status lamps in hard-wired operation panels.

Label Control: can be connected directly to a specific data point in an ABB control. It
can also be used to display status on digital/analog signals. The Label Control can be
configured via property settings to automatically update itself based on the current
state of the communication link to the robot control and the state of the digital signal
assigned to the label control.

In addition, there are two other controls used in the implementation. The RobotBox Control
provides a set of controls that can be used to build user interfaces which can set and display
information from an ABB Robot Controller. Common functions using The RobotBox
Controls are Start-, Stop-, or Step program execution. Finally the ABB FileManager Control
was used to provide a Windows-style file browser for both robot controller disk and PC disks

104
 System Integration

on the network. In most cases, file transfer to the robot was done automatically since only one
generic program was used in the robot controller.

6.1.2 Generic Robot Programming Using WebWare

From this rather broad introduction to the WebWare SDK platform and its underlying
controls, this section will present some of its use for the controller aspect. In section 9.2,
process-oriented simulation is presented including how task-modules are created and how
they can be executed in any order the user prefers. One solution is to send a sequence of
RAPID programs to the robot controller depending on the sequence at hand. The research
within this dissertation solved it differently. The robot controller only stores one small single
program in order to make the robot execution faster. It was concluded early in the project that
the bottle neck within the integration would be the robot row-by-row program interpretation.
Instead it was decided to create one small program and direct the cpu demanding operations
to the PC. The robot RAPID program was kept unchanged; hence it was called a generic robot
program.

Generic programming is about generalizing software components so that they can be easily
reused in a wide variety of situations. One example of components simplifying this concept is
by using classes and function templates in C++, as a way to make the generalization possible
without sacrificing efficiency.

This method of using a generic programming approach enables the sequence of operations to
change dynamically. The process-oriented programming approach, presented in Chapter 9,
executes an operation list, containing rows, where each row is a process. It could be manual
operations, robot operations, measuring operations etc. The user is able to choose any one of
these operations. This approach of generic programming is the support to that approach.

To show an example of communication using RAP, Figure 6.2 shows how execution between
the servo controller and the Master PC is communicating with the S4 controller in the robot.

105
 System Integration

PC computer S4 Controller
ProcessOneTarget() Module A_interr
Select Case Item.Task PERS robottarget RobotPosition:=[[0,0
Case OperationType.FlyBy 1
Call SetPositionData(Item.Target) TEST NextAction
2
Call SpeedData.SetSpeed(Speed) CASE 0:
Call StartRobotMovement(Item.Target) !Idle
While RobotMotionMode=IsDrivingInRobCoor 3
DoEvents CASE 7:
Wend NextAction:=0;
Select Case MoveJSync RobotPosition Speed, fine, tool1, Talk;

End Select ENDTEST
End Sub ENDPROC
4
Private Sub Helper1_SCWriteEvent(. 5 PROC Talk()
If S4ObjName=CurrPos Then CurrPos:=CRobT();
Targ=CreateS4RobTarget(RetValue.List(0),..) SCWrite\ToNode:=130.236.35.16,CurrPos;
ENDPROC
End If RETURN;
ENDMODULE
End Sub
TCP/IP
Figure 6.2: execution between the servo controller and the Master PC

The sequence is:

1. The functions SetPositionData and SpeedData.SetSpeed in the WebWare application

set the S4Target structure in the robot. This specifies where the robot is instructed to
go in the Cartesian robot coordinate system. RAP is updating the variable in the robot
controller via the TCP/IP connection.
2. The only Main program running in the S4 controller is a Case clause. If the integer
NextAction is set to zero the robot is in idle mode (CASE 0). The procedure
StartRobotMovement in the PC Computer is updating the integer NextAction to the
value 7 and RAP is updating the variable in the robot controller.
3. When NextAction is set to 7 the Case clause will execute a MoveJSync instruction,
which moves the robot to the Cartesian robot coordinate RobotPosition. The robot
program interpretation will stop on this line until it reaches the coordinate. When the
robot has reached the location the sub-routine Talk is executed
4. The Sub-routine Talk stores the robots current location in a variable CurrPos and
sends it using RAP to the computer on ip-address: 130.236.35.16.
5. When CurrPos is updated in the Master PC, the event Helper1_SCWrite is executed
and the busy wait loop (While RobotMotionMode=IsDrivingInRobCoor) in the
procedure ProcessOneTarget is released and the sequence of operation can proceed.

106
 System Integration

All the functions in the generic robot program include the following events on NextAction:

CASE 0: is the idle, hence robot does nothing.

CASE 1: uses the RAPID command MoveLSync, which moves the robot to a coordinate in
linear movement. This means that the robot moves on straight lines. One example is just
before docking and at the undocking movement. Due to the high tolerance in the Capto
interface, the robot must move straight to avoid clash.

CASE 2: uses the RAPID command MoveJSynch, which is similar to MoveJSynch. The only
difference is that the MoveJSynch command moves the robot to the coordinate without taking
linear movements in consideration. One advantage with this is that the robot avoids
singularities. One example of a singularity is when joint 4 and joint 6 is in the same line,
which stops the robot program execution.

CASE 3: uses the RAPID command MoveAbsJ, which is forward kinematic function. This
means that the user can set a joint value on all individual joint. This is especially convenient if
the robot is caught up in a singularity. The only way to move out of this mode is to either jog
the robot manually or using the MoveAbsJ command.

CASE 4: uses the RAPID command SoftAct, which makes the robot compliant. In certain
situation, such as undocking it is good to have the robot a bit flexible. This avoids jamming
effects. This command is set to each joint individually.

CASE 5: uses the RAPID command SoftDeAct, which removes the SoftAct setting. After the
robot has undocked a Capto Interface this command is used.

Another advantage to this approach of generic programming has another beneficial

consequence; the robot can be jogged manually from the operation panel on the PC-computer
screen. The user selects a coordinate system to move the robot or joint-wise movement. A
sub-routine in the PC calculates the step, and the MoveL for linear movements, or MoveAbsJ
for joint-wise movement could be initiated in the CASE clause in the generic program
presented above. It would be possible to solve this by generically creating RAPID clauses and
convey these to the robot. But from a scientific point of view, it was considered better to have
as much functionality as possible using the Visual Basic program on the PC Computer. The
intention being, if another robot type is used instead of ABB Robots, the only additional work
required would be to create another generic program for that particular robot type. However,
to accommodate this deficiency a protocol would need to be implemented between the general
Visual Basic program and WebWare, so that a different embedded system from another robot
type could be incorporated into this system. The initial proposal supports that approach, but
from this dissertation perspective, further work is still required.

6.1.3 emScon For Tracker Integration

Similar to WebWare accessing the robot, the tracker-programming interface emScon was
used to integrate the Leica tracker. The Tracker Controller Plus comprises of two computers:
one for the tracker controller, and one for the emScon webserver. The tracker controller
computer samples measurements from the tracker unit and sends the data through a twisted
RJ45 cable to the emScon computer that distributes the data further to a client on the TCP/IP
network. Unlike WebWare, everything that is possible with the laser tracker is accessible
through standard TCP/IP networks. The emScon uses socket communication where a client
on the TCP/IP network sends a request to the emScon server and the server answers back to
the requested IP address. The user has two ways of implementing the emScon interface, either
107
 System Integration

synchronously or asynchronously. Synchronously, the program stops and does not continue
execution until the emScon server has responded upon a request. Asynchronously, an event
handler (objects) is raised to deal with the request.

PC computer emScon webserver

Dim ObjConnect As New LTConnect

Public WithEvents ObjAsync As LTCommandAsync

Set ObjAsync = ObjConnect.ILTCommandAsync

ObjConnect.ConnectEmbeddedSystem 192.168.0.1

ObjAsync.SetContiniousTimeModeParams LaserInterv
ObjAsync.SetMeasurementMode / 1
ES_MM_6DContiniousTime 2 LTConnect

Private Sub ObjAsync_ContiniousProbeMeasData.. 3

ObjConnect.ContintiousProbeDataGet( Data, i,

Set LaserTarget=CreateS4RobTarget(dval1,dval2

End Sub
TCP/IP

Figure 6.3: An example of the asynchronous communication between the Visual Basic
integration software

The interface to emScon consists of a COM (Component Object Model) object. A COM
object is a DLL library and not an include file for C++ or an ActiveX (OCX) control for
Visual Basic. This is in contrast to the robot interface ABB WebWare. It is designed as an
ATL (Active Template Library) type COM server and can also be used in any non-Microsoft
Windows system. LTConnect.dll is the driver enabling the tracker server (SDK) with a built-
in TCP/IP communication. The LTControl COM-object driver is based on the tracker server
TPI (Tracker Programming Interface) the Win32 Sockets 2.0 API. The LTControl.dll is, in a
sense, a tracker server client allowing design of such control. It is not permitted for users to
have direct access to the TCP/IP communication libraries or System Programming Interfaces,
but the high-level TPI supports both synchronous and asynchronous methods.

Shown in Figure 6.3 above, the tracker implementation begins by declaring the ObjConnect
object, which in turn, creates the interface to the LTControl driver. The ObjAsync object is
created as type asynchronous communication for the LTControl interface. The user may have
declared a synchronous object for the interface here. Having declared the objects needed, the
initialization to the tracker is engaged by opening the embedded handler to the emScom
webserver on the IP-address (192.168.0.1 seen in Figure 6.3above).

108
 System Integration

Figure 6.3, shows an example of the asynchronous communication between the Visual Basic
integration software created in this research. The sequence is as follows:

1. By calling the procedure SetContiniousTimeModeParams on the ObjAsync object, the

user declares how many measurements per second are sent to the LTContol driver
from the emScon webserver.
2. By calling the procedure SetMeasurementMode to ES_MM_6DContiniousTime, the
emScon webserver will start transmitting measurements to the LTControl driver on
the requesting computer.
3. Every time a measurement arrives on the LTControl object, the procedure
ObjAsync_ContiniousProbeMeasData will be initiated. This will happen in the
background of the program execution of the Visual Basic integration software. This
procedure puts values from the measurement into global variables that can be used in
the program execution.

6.2 System Topology

6.2.1 Network Topology

The integration software brings together WebWare and emScon into one unit. Even though
WebWare uses ActiveX, and emScon uses the LTControl COM object, the implementation is
similar and goes hand-in-hand. As the robot reaches a location through RAP, the SC_WRITE
event is raised. If a measurement is delivered from emScon, the
ObjAsync_ContiniousProbeMeasData is communicated through the TPI.

WebWare

emScon

Figure 6.4: The network topology of the Integration

109
System Integration

Figure 6.4 shows the topology of the system. The process-oriented program is generated in
the Offline Programming System, as presented in Chapter 9. The XML code is interpreted by
the Visual Basic program. When a measurement routine is requested, the Integration Software
makes a call using TPI. If a stream of measurements is selected, the emScon webserver will
use TPI to return back measurements at regular time intervals. Figure 6.4, also shows how the
Integration Software creates the robot requests via TCP/IP using WebWare and RAP.

6.2.2 System Overview

To give an overview of how the Cartesian space is structured, this section presents the
coordinate systems and transforms in the integration. The mathematical equation to calculate
these transforms uses the notation presented by Craig (1989) and a short summary to that
notation is presented in Appendix C. Figure 6.5 shows an overview of the coordinate systems
and the transform in the integration.

Tr
T
6 D Pr ism
{Tr}

Tr
Rob T Rob
T 6 D Pr ism
TCP 0
TCP 0T

{6DPrism}
{TCP0}
{EETCP}
{Rob}

6 D Pr ism
EETCP T

Figure 6.5: The Coordinate systems and transforms for the integration

The base coordinate system for the tracker is named Tr. All measurements of the reflector
prism are from Tr to the laser target prism, called the 6DPrism. A measurement given by the
tracker is from coordinate system Tr to coordinate system 6DPrism and is the same as the
Tr
transform 6 D Pr ism T . If the position of the robot was asked, the robot would reply the position of
the distance between the coordinate system Rob to the coordinate system EETCP (End-
Effector TCP), which is the repositioned TCP from TCP0 (tcp zero). Without a tool defined
in the robot, the robot gives the positions of its TCP0. By creating a new end-effector tool, the
TCP can be moved from TCP0 to EETCP, hence the transform 6 D PrTCP ism
0T . Within the

Integration Software implemented in this research, all measurements were obtained within the
coordinate system Tr. Also, a transform has to be used in order to acquire all the robot
Tr
positions that by default are given in the coordinate system Rob. This transform is called Rob T.
The calculations to achieve these transforms are presented in Section 7.3. To close the
kinematic chain, an additional transformation 6 DEETCP Pr ism
T is required and can be described as
going from the 6D-Reflector to the Chuck coordinate system.

110
 System Integration

6.2.3 Robot Chuck End-effector

This section also presents the coordinate systems and transforms of the Robot Chuck end-
effector without deriving the mathematics. The mathematical derivation is presented in the
next chapter (Chapter 7). Figure 6.6, gives an overview of the system and is represented by
different colors. The red color is the coordinate system. Blue is the measured location during
the calibration process. The green color is the transforms resulting from the calibration work.
These two transforms are generated by the calibration function presented in section 7.3. The
orange transforms where calibrated using a LTD800 laser tracker supplied by Leica.

111
 TR

{Tr}
C2

Tr
TCP 0

Tr
T {TCP0} TCP 0
T
6 D Pr ism
EETCP {EETCP}

Tr
RobT
TCP 0
T
6 D Pr ism
6 D Pr ism
EETCP T
Rob
TCP 0 {6DFemale} 6 DFemale
T
6 DMale

{6DMale}
6 DFemale
T
6 D Pr ism

Green: Robot Calibration 6 D Pr ism

T
Blue: Measured {Rob} 6 DMale

Brown: Calculated
Red: Coordinates
Orange: Laser Calibrated
{6DPrism}

Figure 6.6: Coordinate systems and transforms of the Robot end-effector

112
 System Integration

The Robot Chuck end-effector uses a Capto chuck to either attach a drilling machine to the robot
or to dock with Dynamic Modules. On both sides of the chuck there are two male Capto
interfaces that are used to attach the 6D-Reflector on each side of the end-effector. This enables
the tracker to have greater access to the robot. Figure 6.6 shows a drawing of the end-effector
and the two coordinate systems EETCP and TCP0 in the x-y plane.

The Chuck is driven by a hydro-mechanical chuck system and is controlled using the available
I/O.

6.3 Example of Probing, Docking and Reconfiguration

Before presenting the calibration methods for making this integration possible, this section is
briefly showing the sequence in reconfiguring a Dynamic Module between two configurations.
The sequence is graphically shown in Figure 6.7a to Figure 6.7f shows a step-by-step procedure
for a docking operation followed by the reconfiguration operation.

113
System Integration

Figure 6.7a: Probing of module Figure 6.7b: Going to dock

Figure 6.7c: Robot docked with hexapod, Figure 6.7d: Moving hexapod with robot
hexapod unlocked

Figure 6.7e: Hexapod adjusted to +/-50 um Figure 6.7f: Hexapod locked and released
and +/-0.0005 rad

114
 System Integration

6.4 The Integration Software

Software within the Master Controller PC is shown in Figure 6.8 and was developed using Visual
Basic to execute the sequence of operation retrieved from DELMIA. Chapter 9 explains in detail
how the sequence was generated using DELMIA. This section is showing how the sequences are
executed in the workshop using the integration software. This section breaks down three of the
most important windows in the integration program.

6.4.1 Control Panel

The Control Panel, called LiU Control, is the top level window that is first shown as the program
initiates. Figure 6.8 shows the LiU Control main window. Each label is referring to a
functionality that is further explained under this subsection.

1 4 7

2 5 8

Figure 6.8: The LiU Control Panel shown on the left and the
Operation Panel shown on the right

The Integration software can perform most operations done with the ABB S4 Teach Pendant.
This is made possible by the generic robot programming approach that was presented in Section
6.1.2. RAPID functions are embedded in scripts that are executed via WebWare and the RAP
(Robot Application Protocol). The Control Panel has the following functionality (the numbering
refers to the numbers in Figure 6.8):

115
System Integration

1. Cartesian Space for output and input of values

a. The operator can read the position of the robot or tracker or any coordinate
selected. Positions can be read either continuously or statically.
b. Operator can input new values to store as coordinates or move to with robot.
2. Transformation between Euler angles and Quaternions.
a. Euler values are more intuitive than Quaternions and are used by the operator to
input/output angle values
3. Accuracy on active coordinate
a. Accuracies are given in the active coordinate system (tracker, robot or any
selected coordinate system)
b. The two yellow color bars show the square root sum of x,y and z and pitch, jaw
and roll respectively.
4. Robot menu
a. Robot I/O, where all the surrounding equipment are accessed and manipulated
b. Maintenance
c. Joint programming, where the operator can manipulate the robot joint vise
d. Velocity menu, which enables setting the general speed settings of the robot
e. Tool menu, which accesses the customization of end-effectors
f. Robot parameters, such as speeds, soft servo parameters are accessible here
5. Dynamic Module menu
a. This menu accesses all functionality to the Dynamic Modules
6. Laser menu
a. Measurement modes on the tracker are accessed here
7. Coordinates
a. Operator can store and load Cartesian coordinates from this section
b. Coordinates are drag-and-droppable between fields
c. Coordinates can be loaded and stored from the disk here as well
8. Coordinate transformation menu
a. This menu handles which coordinate system is to be active
b. By drag-and-drop coordinates from the Coordinate Menu, different transforms are
used. The transforms are multiplied with the current.
c. More than one transform can be dragged into one field

6.4.2 The Operator Panel

By clicking the Import button in the Operator Panel, the sequence of operations is imported
from the offline program, created in DELMIA. The offline created program in DELMIA is stored
in an XML based file. The XML file only has the sequence of operation and no ABB RAPID
code. This is explained in detail in Section 9.2. The operator can select any operation in the
operation list and the integration software starts the execution at the selected line. The Operator
also has the option to execute in automatic mode.

The Operation Panel contains the following functionality:

Import XML-code from local hard drive or from a shared computer, i.e. the OLP
computer
Click on any command in the operation list and execute either step-by-step or automatic
Load/Save the list

116
 System Integration

The sequence of operations, contains both manual and robot operations. If the list executes the
operation ProbeModule, the Probing Menu is popped up.

6.4.3 The Probe Menu

The Probe Menu appears when the task ProbeModule is selected in the operation list and a new
menu pops up for the operator (Figure 6.9). This menu presents the CAD data to the user. This
gives the user the possibility, not only to measure the location of the Dynamic Module, but to
position it correctly and match the nominal CAD position as close as possible. When the operator
has repositioned the Dynamic Module within centimeter tolerance, the status Good is displayed
in the Measurement Status of the Probe Menu. Figure 6.9 shows the Dynamic Module is within
tolerance for both x and omega. The sequence is the following:

1. Clicking the button: Measure Continuous will update the values in the Measure Status
section. Once the Dynamic Module is within tolerance the Good value is set
2. When all six degrees of freedom are set the Dynamic Module is locked and the button
Stop Continuous
3. This button changes the tracker to measure in stationary mode
4. Clicking the button: Stationary Mode performs an accurate measurement over a period
of two seconds.
5. Update Location updates the stored tag frame to the new values. The operation FlyBy
before DockingModule and the DockingModule itself will be updated from this
measurement.

Figure 6.9: The Probe menu for roughly position Dynamic Modules
and to probe their initial location.

117
System Integration

As the operator has finished measurement by clicking the button Update Location shown in
Figure 6.9, the program in the Operation Panel will continue executing the next operation in the
operation list.

6.5 Metrology Feedback Controller Calculations

When reading this section the reader is recommended to look in Figure 6.6, which gives an
overview of the coordinate systems and transforms. When the Master Controller (Figure 6.4) is
running, a feedback loop measuring is executed. The feedback loop measures the actual location
of the robot and the controller calculates the error between the nominal value (tag frame from
CAD) and the actual value. Measurements are done using the 6D-Reflector; hence that
measurement is transformed into the coordinate system of the chuck end-effector attached on the
robot {EETCP}. The transform calculation is:

Tr 6 D Pr ism
EETCPi = Tr 6 D Pr ismi EETCP T,

where Tr 6 D Pr ismi is the measurement of the 6D-Reflector and the transform 6 DEETCP Pr ism
T is
transforming the measured value to the chuck of the robot. Prior to the error calculus the point
value (x, y, and z) is extruded from the frame. A frame object in the integration program is
actually a seven length vector containing x, y, z, q1, q2, q3, q4, where the q-values are the
Quaternions that are describing the orientation of the frame. In the integration program these
vectors are called targets.

6.5.1 Translational Error

For calculating the translational error between two targets, the x, y, and z values are extruded
from the frame target and the total error distance between two targets can be calculated.

Z Nom {nominal}
{actual}
Tr
Z Act X Act Err
X Nom
Tr
Tr YNom EETCPNom
EETCPAct
YAct
The author calls this error calculus absolute value of the positional error, and is calculated by:

Tr
Errxyz = (Tr X Nom Tr X Act ) 2 + (Tr YNom Tr YAct ) 2 + (Tr Z Nom Tr Z Act ) 2 .

But what about the orientation? Well, the orientation error calculus is actually incorporated to the
feedback loop.

118
 System Integration

6.5.2 Orientation Error

Because of the non linear properties of rotations a mean value is not a simple add and divide by n
operation. For example when the angle between two rotations in a calculus is 180, an infinite
number of solutions will exist. One way to calculate the mean value is to use spherical linear
interpolation of Quaternions (SLERP). SLERP is a method of interpolating two rotations
smoothly. It is called spherical-linear since the two quaternion rotations are interpolated
uniformly along a geodesic in the surface of a 3D-sphere. Assume the orientation error
corresponds to the angle between q Act and q Nom , see Figure 6.10.

q Act Interpolation is done

along this arc
q Nom

Figure 6.10: The orientation error is the difference between q Act and q Nom

Err in Figure 6.10 corresponds to the difference between the actual value of the orientation and
the nominal value of the orientation. Using the SLERP method the Err is solve by calculating the
vector dot product between q Act and q Nom according to:

q0Act q0Nom
q1Act q1Nom
cos( Err / 2) = ,
q2Act q2Nom
q3Act q3Nom

which can be rewritten as:

q0Act q0Nom
q1Act q1Nom
Err = 2 arccos .
q2Act q2Nom
q3Act q3Nom

From this equation Err gives the error in radians.

6.5.3 Termination Criteria

This calculation corresponds to the feedback loop in the Metrology-integrated Control. The
feedback loop stops iterating when both the translation and the orientation error are within the
specified requirements. Typical threshold for the translation error ( Tr Errxyz ) has been 0.05 mm
(e.g. 0.05 mm) and for the orientation error ( Err) has been 0.0005 rad, which is close to 0.03.

119
System Integration

All calculations are done in the Tracker coordinate system to avoid transformation errors. A new
value is sent back after each loop cycle to the robot for correction, which is:

Tr
Rob
EETCPi +1 = Rob T ( Tr Err + Tr
EETCPi ),

Tr Tr
Where Err contains both Errxyz and Err. This means that the robot is not running on the
nominal frame value, but on a modified value by the Master Controller. Appendix A shows two
plots for the feedback loop when reconfiguring a Hexapod to high accuracy. Figure A1 shows
how the error is reduced to within a tolerance interval that is set to 0.05 mm and Figure A2
shows how the orientation is reduced to within 0.03. Note in the figure that the x, y, and z
values for the robot is far from the nominal value.

120
 System Setup Calibration

7 System Setup Calibration

This chapter is presenting the transform equation between coordinate systems and frames that is
calculated in different calibration procedures. This chapter will also present calibration
procedures that are performed once in the setting up of the system and the calculation performed
in the probing procedure that is done before the robot reconfigure the Dynamic Modules.

When following the derivations and explanations of the calculations, it is advised to look in
Figure 6.6, which shows the transformations between the different coordinate systems of the
integration.

7.1 6D-Reflector Calibration

In Section 5.2.3, the conceptual principle and application of the 6D-Reflector was presented. This
Section presents the calibration method that makes the 6D-Reflector work in practice. As
mentioned previously, the Capto interface was used to attach the reflector to the robot or tooling.
The tracker measures the location of the prism, called 6DPrism. To measure the female and male
interface of the reflector, the two transforms 66DFemale
D Pr ismT and
6 D Pr ism
6 DMaleT are required. To find the
transform during the calibration, the reflector was attached onto a rigid Capto chuck.

Conceptually, having installed the Capto interface to the 6D-Reflector, a LTD500 Laser Tracker
was used for the calibration. All these measurements were carried out using the software Axyz,
which is a dedicated Leica Tracker software. The measurements were taken at Leica Metrology
R&D center in Aarau, Switzerland. The conceptual idea was to:

1. Measure the machined surfaces on the male and female surfaces of the Capto interface.
These surfaces are Z=0 for 6DMale and 6DFemale.
2. Measure the circle center of the male and female Capto interfaces. The circle center is the
center of the Z-axis.
3. Measure a machined hole pointing in radial direction on the male part. This hole will
direct the X-axis.
4. Create the two coordinate systems 6DMale and 6DFemale. The coordinate systems are
calculated using the Leica Axyz software.
5. Measure the 6D-location of the 6DPrism.
6. From the three coordinate systems 6DPrism, 6DFemale and 6DMale, calculate the
transforms 66DFemale 6 D Pr ism
D Pr ismT , 6 DMaleT and
6 DFemale
6 DMaleT .

121
System Setup Calibration

Step 1-6 above generates three coordinate systems, as seen in Figure 7.1. 6DPrism is the actual
laser target that the tracker performs its measurements. 6DFemale is the female part of the Capto
Interface (Figure 7.1), while 6DMale is the male part of the Capto Interface.
6 DFemale
T
6 DMale

{6DMale}
{6DFemale}
{6DMale}
{6DPrism}
6 DFemale 6 D Pr ism
T
6 D Pr ism 6 DMale T

{6DPrism}

Figure 7.1: The Capto interface as part of the 6D-Reflector

The Capto system has a repetitive accuracy of 2 m, so the coordinate system, that either the male
or the female Capto interface is attached to, is assumed to be aligned with the 6DFemale or
6DMale part. This of course depends on what the reflector is attached to.

7.2 Robot Chuck Calibration

Having calibrated the 6D-Reflector, it can then be used to calibrate the Robot Chuck. When
running the integration program, such that {EETCP} is positioned with high accuracy, we need to
calculate the position of the chuck by using a transform. A measurement with the tracker is stored
in the frame Tr 6 D Pr ism n, where n is the index on the measurement. All measurements and
calculations are performed in the Tracker coordinate system {Tr}. The 6D-location of {EETCP}
is given by:

Tr
EETCPn = Tr 6 D Pr ismn 6 DEETCP
Pr ism
T.

6 D Pr ism
Figure 7.2 shows graphically how the transform EETCP T can be derived.

122
 System Setup Calibration

Tr
6 D Pr ism A
6 D Pr ism
C2 6 DMale T

6 D Pr ism
EETCP T

{EETCP}
C1

{Tr}
Tr
6 D Pr ismB

6 D Pr ism
Figure 7.2: The required transforms to calculate EETCP T

The step-by-step method to perform the calibration is:

1. Attach the 6D-Reflector onto the chuck using the male Capto;
2. Measure the 6D coordinate with the trackers 6DPrism, named frame Tr 6 D Pr ism A ;
3. Detach the 6D-Reflector and reattach it at location C1 or C2 (next to the Capto chuck);
4. Measure the 6D coordinate with the trackers 6DPrism, named frame Tr 6 D Pr ismB

6 D Pr ism Tr
Calculate the transform EETCP T using the two measured frames 6 D Pr ism A , and Tr 6 D Pr ismB .
6 D Pr ism
Deriving the transform EETCP T for the coordinate system EETCP, the transform equation is:

Tr 6 D Pr ism Tr 6 D Pr ism
6 D Pr ism A 6 DMaleT= 6 D Pr ismB EETCP T.

By multiplying both sides with Tr

6 D Pr ismB -1 the transform 6 D Pr ism
EETCP T is calculated from:

Tr
6 D Pr ismB -1 Tr 6 D Pr ism A 6 DMaleT = Tr 6 D Pr ismB -1
6 D Pr ism Tr
6 D Pr ismB 6 D Pr ism
EETCP T,

which becomes:

6 D Pr ism
EETCP T = Tr 6 D Pr ismB -1 Tr 6 D Pr ism A 6 D Pr ism
6 DMale T.

The transform 6 D6 DMale

Pr ism
T is known from the calibration presented in Figure 7.1 (note that the inverse
was needed to invert the direction). The other measurements Tr 6 D Pr ism A and Tr 6 D Pr ismB are
provided by the tracker from the two measurements in step 2 and 3 above, hence the transform
6 D Pr ism
EETCPT can be calculated.

123
System Setup Calibration

The same assumption, as described in Figure 7.2, shows that the coordinate system {EETCP} is
aligned with the coordinate system 6DMale. This is because of the high repetitive accuracy in the
Capto tool changer. However, it is important to keep the metrology end-effector absolutely still
during the calibration procedure (attaching and re-attaching the 6D-reflector). It is therefore
recommended to remove the robot end-effector and rigidly attach it using a bench during this
calibration phase. When the chuck position is given from the tracker, the calibration is the only
process that will affect tolerance build-up in the positioning of the robot. This statement is further
explained in the next section.

7.3 Robot Base and TCP0

Tr
The calibration to generate the transformations between the Tracker and the robot ( Rob T ) and
between TCP0 and 6D-Reflector ( 6 D TCP 0
Pr ismT ) is a simple approach. The method is based on moving

the robot to several locations. Once stopped each location is stored in both the tracker measuring
the 6D-Reflector and the TCP0 locations in the robot base coordinate system. All locations are
stored in a text file that is transferred to an optimization routine within Matlab. This routine has
twelve unknown parameters to be solved that constitute the two transforms. Therefore, to make
this optimization work, twelve measured locations are needed. Furthermore, this section
describes the procedure used for this calibration.

Tr
The first transform needed to go between the Tracker and the robot, is the transform Rob T . The
relationship between the coordinate systems {Rob} and {Tr} is represented by:

Tr
{Rob} = Rob T {Tr}.

With respect to all measurements carried out in the integration software, the calculations are
always performed in the tracker coordinate system {Tr}. Hence, the transform needs to go from
Tracker coordinates {Tr} to Robot coordinates {Rob}. The inverse of the transform is taken
according to:

{Tr} = T 1 {Rob}.
Tr
Rob

In order to calculate 12 unknown entities in the two transforms, at least 12 measurements had to
be stored within a text file, which gives the position of the TCP0, in Robot coordinates {Rob},
from each position of the Prism in Tracker coordinates {TR}. The representation of a measured
TCP0 location of the robot is given by RobTCP. The equivalent stored location of the Prism in the
Tracker coordinate is given by TR6DPrism. The transform equation for each measurement is
derived from Figure 6.6, and becomes:

Tr
6 D Pr ism=Tr TCP0 6 D TCP 0
Pr ismT

To further explain the relationship between coordinate systems, let us now go via coordinate
system {Rob} to coordinate system {TCP0}, instead of going from coordinate system {Tr}. First
we see in Figure 6.6 that:

Tr Tr Rob
TCP 0 = Rob T TCP 0 ,

124
 System Setup Calibration

which we could define:

Tr Tr Rob TCP 0
6 D Pr ism = Rob T TCP 0 T.
6 D Pr ism

Furthermore, we can move over all entities to one side:

Tr Rob TCP 0
Rob T TCP 0 T Tr 6 D Pr ism =0
6 D Pr ism

Rob
By moving to at least 12 different locations, storing each location TCP0 and Tr6DPrism, the
transform equation can be re-written according to:

12

Pr ismT 6 D Pr ismi = 0 .
Tr
T RobTCP 0 i 6 D TCP
Rob
0 Tr

i =1

After the calibration, RobTCP0 and TR6DPrism were constructed out of variables using the
optimization function: fmincon, included in the MATLAB Optimization Toolbox. The
optimization is performed by minimizing the cost function:

f = Tr
Rob T Rob TCP 0 i 6 D TCP
Pr ism T
0 Tr
6 D Pr ismi
n

Note the index i above represents the measured pair of values for the robot and Tracker. The
fmincon finds the constrained minimum of the function containing several variables. The
Tr
function responds with the two transforms Rob T , 6 D TCP 0
Pr ismT and the minimized value. Several tests
showed the value of the cost function being low, and further testing indicated that the robot
coordinate system aligned with the tracker coordinate system.

As mentioned earlier, it was stated that the quality of this calibration is surprisingly does not
affect the absolute accuracy in the robot positioning during program execution! An explanation
for this is that since all measurements are done in the tracker coordinate system by the tracker,
the controller will make sure the robot is in the correct position. Hence, poor quality within this
calibration will only lead to further iterations of the controller loop. However, an improved
calibration will allow the robot to position to higher accuracy locations faster, with less loop
cycles in the controller.

7.4 End-effector calibration

When manipulating a robot with end-effectors, the TCP of the robot is often moved from TCP0
to the end-effector tip. This enables the robot to rotate around the tip of the end-effector unit.
Normally the TCP coordinate system is moved to simplify manual jogging of the robot on the tip
of the end-effector. By this research, two techniques have been defined for performing end-
effector calibration. The first method is to calibrate online, while the second method is offline.

7.4.1 Online calibration of End-effectors

This method is based on the scenario that the end-effector is already attached to the Robot Chuck
and calibration of the end-effector tip is to be calibrated. The tip is interesting to calibrate,
especially for drilling machines, because the tip is the bushing that is pressed on the surface to be
125
System Setup Calibration

drilled and decides the accuracy of the drilled hole. In this section the tip of the end-effector to be
calibrated is simply called TCP. For this calibration, the robot needs to be standing still and the
production process is stopped. This calibration is most useful if done regularly, such as adjusting
a pressure foot. The relationship between each coordinate system is presented in Figure 7.3:

{TCPDrill}
Tr
TCPDrill Prism

{Tr} 6 D Pr ism
T
TCPDrill
Drill bushing

Tr
6DPrism
{6DPrism}
Figure 7.3: Calibrating an end-effector online

The method of the calibration is summarized in the following steps:

1. Measure the plane of the pressure-foot using a reflector ball. The measurement generates
the XY-plane.
2. Measure a cylindrical object in the same direction as the drill on the machine. In this case
it is not important that the cylindrical center is aligning with the drill Z-axis. This
measurement generates the Z-axis direction
3. Position the reflector ball in the drill bushing. The reflector ball does not need to be on the
XY-plane. The measured point is simply transformed to the XY-plane. This measurement
generates the location of the Z in the XY-plane.

The transform cannot be measured directly but TrPTCP and TrP6DPrism are measured locations. The
transform equation is:

Tr
TCPDrill= Tr6DPrism 6 D Pr ism
TCPDrillT

6 D Pr ism
whereby the transform TCP T is calculated according to:
6 D Pr ism
TCPDrillT = Tr6DPrism -1 TrTCPDrill

7.4.2 Offline calibration of End-effectors

Probably the most reliable and obvious method to calibrate the TCP of an end-effector is to
measure the unit in a CMM (Coordinate Measurement Machine). A CMM has an absolute
accuracy in the order of 10 m, and hence, presumably better than a hand-held reflector ball.
Essentially, the same transforms as the online calibration method can be used. The difference
being the probe of the CMM will measure the Capto interface of the end-effector. Since this
Capto interface is where the robot holds the end-effector, the transform to be solved, using the
offline- calibration method, is Chuck
TCPT . This can be seen in Figure 7.4. This transform is then used

126
 System Setup Calibration

by the robot controller as the new tool. The Offline method is more time consuming than the
Online method, but provides higher accuracy because of the CMMs more stable environment.

Chuck

Chuck
TCPT

Figure 7.4: A drilling end-effector

Which one of the two methods is selected, depends on the production scenario. If time is
available and higher accuracy is required, then the offline method is appropriate. In some cases
the already available metrology system provides sufficient accuracy to calibrate and end-effector
and negates any wasted time using a CMM. It is a good role of thumb to calibrate in much higher
accuracy than what is needed. Tolerance build-up is always raising fast.

7.5 Probing of Dynamic Modules

As mentioned in Section 7.3, calibration of the robot is carried out after it has moved, or in some
cases if the tracker unit has moved. However, it is assumed that we do not need to move robot or
tracker when we change to a new product within the ART system. Therefore, a changeover to
another product type is to be carried out in the ART system. The only calibration needed in this
process is the Probing. Probing identifies the positions of the Dynamic Modules (DMs),
presented in Section 4.3.2. Note that if the DMs are only to be reconfigured, and hence not
moving the modules manually, the robot would have previously remembered each of their
positions. If on the other hand, we need to lift in a new model, or the work volume within the
DMs is exceeded, a probe operation becomes necessary.

7.5.1 Probing Using a Non-calibrated 6D-Reflector

At an early stage of this research, the 6D-Reflector was not calibrated. This means that 6DMale
and 6DFemale was not calibrated. However, it was still possible to probe and dock with the robot
in the probed position. This section will present a method to do so. The idea is to generate a
virtual coordinate for the robot to dock onto, and hence, not the real coordinate on the top of
the DM. The virtual coordinate is the actual location of 6DPrism when the reflector is attached to
the DM during probing. Before presenting the step-by-step method to do this, Figure 7.5 shows
the coordinate space.

127
System Setup Calibration

Tr
6 D Pr ismDocked {6DPrismDocked}
{Tr}

Tr
6 D Pr ismPr obed C1
Docked
Pr obed T TCP 0
T
Docked

Tr
TCP 0
T {TCP0}
Rob T Pr obed

Rob
TCP
C2
{6DPrismProbed}

{Rob}

Figure 7.5: The coordinate systems and transforms for non-calibrated 6D-Reflector
Tr Tr
From Figure 7.5 above, the values 6 D Pr ism Docked , and 6 D Pr ismPr obed , are measured with the
Rob
tracker where the value TCP is the position of the TCP0 in the robot coordinate system and
Tr
the known transform Rob T was calibrated earlier in Section 7.3. The unknown transforms solved
TCP 0
in the following step-by-step method are, DockedT , PrTCP 0
obedT and
Docked
T.
Pr obed

The method of calibration is described by the following:

1. Dock the end-effector to a rigidly attached male Capto interface

2. Measure Tr 6 D Pr ism Docked
TCP 0
3. Calculate T from:
Docked

Tr Tr Rob TCP 0
6 D Pr ism Docked = Rob T PTCP 0 Docked T , hence

TCP 0
DockedT= Tr
RobT 1 TCP 1
Rob Tr
6 D Pr ism Docked

4. Undock the robot and attach the 6D-Reflector to the male Capto interface
5. Measure with tracker Tr 6 D Pr ismPr obed

128
 System Setup Calibration

TCP 0
6. Calculate T from:
Pr obed

Tr Tr Rob TCP 0
6 D Pr ismPr obed = Rob T TCP Pr obed T , hence

TCP 0
T=
Pr obed
Tr
Rob T 1 Rob
TCP 1 Tr 6 D Pr ismPr obed ,
Docked
7. Calculate Pr obed T from:

TCP 0 TCP 0 Docked

T=
Pr obed Docked T T , hence
Pr obed

Docked
Pr obed T= TCP 0
Docked T 1 PrTCP 0
obedT .

The 6D-Reflector is used as a probe so that the robot can learn the location of the Dynamic
Modules. This is done prior reconfiguration, before docking the robot to a Dynamic Module. To
inform the robot the location of the Dynamic Module, the only calculation required is:

Rob
TCP = Tr6DPrismProbed TCP 0
Pr obed T 1

In addition, by using the Metrology-integrated robot control to reach a desired high accuracy
location, the system continues iterating until the tracker measurement is within the tolerance of
the following frame value:
Tr Tr
6DPrismDocked = 6DPrismProbed Docked
T 1
Pr obed

This calibration procedure was adequate for preliminary docking operations. However, the tag-
points (coordinates), transferred from DELMIA using the offline programming method, were
defined in the End-effector Chuck coordinate system. This brings us further to the next
subsection.

7.5.2 Probing Dynamic Modules using a calibrated 6D-Reflector

When all tag points are generated in the OLP software, all tags are defined in the robot base
coordinate system {Rob}. In the execution of the integration program, presented in Section 6.4,
all points are located in the tracker coordinate system. Subsequently, all tags need to be
Tr 1
transformed by Rob T . The probing brings all tags from the robot coordinates {Rob} to tracker
coordinates {Tr}, but also updates these tag values from the probed value. Interestingly, when the
tag value from CAD data gives the operator an indication of where each particular DM is to be
relatively positioned, the probing operation updates that tag point. This method is presented from
an operator perspective in Section 6.4.3. From this point the probing is carried out using the
female attachments of the 6D-Reflector, and all probed Dynamic Modules are measured and
stored in TrDM.

129
System Setup Calibration

The measurements are calculated from:

Tr
DM = Tr6DPrismProbed 6 DFemale
T
6 D Pr ism
-1

The transform 66DFemale

D Pr ismT is known from the Figure 7.1, so the calculation is simply the inverse of
the transform such that the direction of the transform is changed.

130
PART IV: OPERATION PLANNING FOR AFFORDABLE AUTOMATION

PART IV
OPERATION PLANNING
FOR AFFORDABLE AUTOMATION

In Part IV, the operation planning methods are presented. Part IV begins with an
introduction to robot programming and simulation in Chapter 8. This is followed by
the operation planning methodologies and programming approach developed within
this research and presented in Chapter 9.

131
132
 Introduction to Robot Programming and Simulation

8 Introduction to Robot Programming and Simulation

This chapter introduces the reader to the simulation and programming used today in the
operation planning of robots and manufacturing systems. This chapter also presents the software
platform used for this research.

Production planning must start before the design of the product is complete. This situation is
difficult since changes within the product design effect the layout of the production equipment.
This is one of the reasons major change in product design close to launch date is expensive.
Therefore, it is preferable to make any decisions on final design of product and production
equipment as early as possible. However, there is a lack of information early in a production
development necessary to support decision making. Looking at the product lifecycle (the x-axis
in Figure 8.1), moving from concept to launch date, the ability to change decreases over time as
more and more decision on product and process design are made (see the curve: ability to change
in Figure 8.1). Implementing late design changes becomes more costly relative to time (see the
curve: cumulative cost of change in Figure 8.1). Simulating performance, construction and design
can be used to reduce the need to build hardware in the early design stages of a new product and
decisions regarding the product can be extended. (Axelsson, 2002).The key point is to minimize
product design changes late in the lifecycle. This can also be applied to the tooling design and
manufacturing process changes. By validating the manufacturing process and resource
requirements early in the design cycle using virtual technology (see the left side of the graph in
Figure 8.1), the program cost has the potential to be reduced. Beside, reducing cost from changes
in the virtual world, due to the ability to predict and act, program launch date can be significantly
shifted to the left in Figure 8.1.

133
Introduction to Robot Programming and Simulation

Cost Cumulative Cost Ability to change

Of Change

Virtual Traditional
Manufacturing Methods

Validation in Validation in
Virtual World Physical World

Ability to
Change

Time
Product Life Cycle Concept to Customer Launch Date

Figure 8.1: Changes in the virtual world are cheaper and easier to
implement than in the physical world

In saying this, 3D CAD & Digital Mock-up (DMU) technologies for product design provide huge
benefits to the design engineer. Unlike the traditional methods, digital mock-ups decrease the
need for expensive physical mock-ups. Designers can model a digital mock-up of a product in
full detail, simulating all functions and anticipate interactions amongst different components.
DMUs contain the complete data associated with the product, and provide superior information
feedback compared to physical mock-ups. This information can also be used to perform
calculations, which would not be possible with a physical mock-up. Simulation systems support a
large range of robotic applications dedicated to specific manufacturing industries. These include
handling mechanisms, arc welding, spot welding, gluing, grinding, painting, riveting, stamping
lines etc.

One software tool suitable for digital mock-up technology is Computer Aided Robotics (CAR).
These are graphical computer programs used for a variety of tasks in production engineering and
programming. The traditional way of programming a robot is by teach-in, where the operator is
moving the robot via a joystick on a teach pendant. This method is generic and relatively straight
forward. As the operator teaches the robot to move to different locations the joint positions are
stored in the robot memory. By later retrieving these joint positions the robot can return to these
locations within its repetitive accuracy. An alternative is to use a second host computer for
programming and is referred to as an offline programming system (OLP). CAR systems can be
used for tasks such as offline-programming; design and simulation of manufacturing systems;
design of robot manipulators and tele-operations (Oscarsson, 2000). Craig (1989) defines OLP as
a robot programming language which has been sufficiently extended, generally by means of
computer graphics, whereby development of robot programs can take place without access to the
robot itself. Craig further explains that offline-programming includes:

Spatial representation of solids and the graphical rendering of these objects

Automatic collision detection
Incorporation of kinematics

134
 Introduction to Robot Programming and Simulation

Path planning
Dynamic Simulation
Simulation of sensors
Concurrent programming
Translation by post processors to various target languages
Workcell calibration

Nikoleris (1991) explains the two most difficult problems with maintaining accuracy in offline
programming are; 1) the work-cell requires being accurate to the virtual model, and; 2) the robot
needs to be as accurate as its repeatability. In reality work-cells differ from the nominal
programmed data and by using calibration modules within the programming environment, these
issues can be resolved. Nikoleris further explains, that the simulation models of robots are more
difficult to calibrate due to each robots individual mechanical structure and as result, each
individual robot has to be calibrated and brought back to the programming environment.

Today, offline programming systems are able to do more than just simulate robot trajectories and
perform assembly simulation. Simulation technologies are also able to model the interaction of
several manufacturing processes, manufacturing resources and product maintenance issues. The
key point is to minimize product design changes late in the product lifecycle. This concept
concerns the whole manufacturing process and not only product changes, but also process
changes.

The advantages of using offline programming systems include:

Developing simulation models increases the understanding of a system

Eventual or existing problems can be tracked down by running different scenarios
Answers to what-if questions can be provided
It is sometimes easier to understand a graphical simulation model in contrast to analytical
approaches especially those having animations
It is an effective training tool
It helps in choosing the best system for a particular application
To comprehend the basic concept of simulation does not require a very deep knowledge
of mathematics
Simulation and programming is done in one system language regardless of the robots
type
The simulated robot movements are used for programming the robot

The disadvantages using offline programming systems include:

In order to build the models the user must have special training and knowledge
Simulation results may be difficult to interpret
Modeling and analyzing the results can be time consuming and expensive
People with good process knowledge are seldom involved with generating the simulation
No guarantee for optimal solution

From an introduction to geometry simulation and robot programming next subsection will
continue a deeper discussion of robot programming.

135
Introduction to Robot Programming and Simulation

8.1.1 Abstraction Levels in Robot Programming

This subsection covers the different level of abstraction for the programming of industrial robots.

Joint-oriented
The first method of programming a robot requires the storing of joint values of each robot joint.
For the robot to move back to a previously visited point, the robot controller establishes the
values of each robot joint with no concern for the position of the TCP (Tool Center Point).
teaching the robot where to move is called Teach-in, or as Craig (1989) calls Teach by
showing. The definition of Teach-in programming is:

The robot itself is used to generate the robot program

From a kinematic point of view, this is what is called forward kinematics. Only the joint values
of the robot are taken in consideration. This programming abstraction level is called Joint-
oriented. Joint-oriented programming has its advantages. Tolerances in the processes are
maintained whilst the robot has good repetitive accuracy to return to a stored configuration. The
absolute positioning is achieved by the eye of the operator. The joint-oriented abstraction level of
programming does not consider coordinate systems. Programming a robot with only joint values
can be cumbersome unless the operator physically holds the process tool attached to the
manipulator when teaching. In most cases it is preferable to physically distance yourself from the
robot for safety reasons and alternatively the next level of programming includes the awareness
of the coordinate systems within the robot system. This brings us further to the next abstraction
level of programming robots.

Robot-oriented
In robot applications the robot is working in an
automation cell in relation to other objects and
fixtures. To define the robots location in a
program it is mostly convenient to work in
coordinate systems that are in accordance with
logical operators, such as lines and circles. This is
called robot-oriented programming. Bolmsj
(1989) defines robot-oriented programming by all
instructions that explain the task is directly related Figure 8.2: Different coordinate
to physical or logical operations in the robot systems (Cartesian targets)
system. This is sometime called explicit
programming (Bolmsj, 1989; Craig, 1989). Programming robots in robot-oriented programming
can still be done according to teach-in, as in the joint-oriented programming methods, but in this
case the operator has access to several more functions in the robot program language. Further
more, other coordinate systems, such as fixture, product or world coordinate systems can be used
when programming. In robot-oriented programming, the robots TCP can be instructed to travel
on linear or circular paths. The TCP is calculated from the inverse kinematics that calculates the
necessary joint angles for each increment moved in the Cartesian space. The robot is instructed to
move to different Cartesian targets, which is a three dimensional point in space (the location in x,
y, and z and the orientation of each axis Rx, Ry, and Rz). The different Cartesian coordinate
systems for robots are presented in Figure 8.2. In robot-oriented programming, robot locations
and references are used to program the robot. Joint angles on the robot are not considered

136
 Introduction to Robot Programming and Simulation

explicitly. Robot-oriented programming methods use the robot to program the robots
movements. This also requires the operator to have an in-depth technical knowledge on how to
build advanced movements and operations. In addition, it is difficult to program multiple robots.
This brings us further to the next level of programming robots.

Object-oriented
Working in a 3D-geometry environment, the programmer of robots does not have to know
about different robot languages. In a 3D environment a universal robot language is used, which
is the same for all robots. In short, the program language is based on defining coordinates in the
Cartesian space, which in this dissertation are called tags or frames. The tags can be used for
different robot types. For example; if the user wants to change a robot model, the same tags are
used without changing anything. The user uses the objects such as tags and modeled geometries
to program the robot. Bolmsj (1989) defines the task, carried out by the robot system in the
object-oriented abstraction level of robot programming, to be explained by the help from
operations on objects and events. As the simulation of robot movements are finished, the robot
model specific code is post-processed and stored in a file that can be downloaded to a robot
controller for program execution. Thus, the robot specific code is basically the result from the
simulation of robot tasks in the virtual robot cell. The object-oriented abstraction level of
programming robots is a kind of offline-programming, but offline-programming can also be
performed on a robot-oriented or joint-oriented abstraction level.

Object-oriented abstraction level of robot programming has been simplified and reduces the time
for the programming of robots. Because the operation planner does not require knowledge of the
specific robot programming languages the learning time for the generation of advanced robot
programs is reduced enormously. In large product volume scenarios this method has proven to
reduce cost at a great extent. In a production scenario where programming is repeatedly changed
or where there is a limited number of manufactured products, the hands-on-work required to
teach the 3D-model locations of the robot is extensive. In this case the abstraction level
programming is not sufficient. The author of this dissertation claims that robots today are still
programmed according to teach by showing, where the operators still show the robot what to
do. However, today, the programmer is teaching a virtual robot, instead of the physical robot.
This brings us to the next abstraction level of robot programming.

Process-oriented
If a scenario would be to plan the robot tasks of the assembly of 30 aircraft wings where each
wing requires the robot to drill 4,000 holes, that would require 120,000 tag points to be
programmed. How could that be efficient using the object-oriented abstraction level of robot
programming? In the wing assembly scenario the CAD-model of the wing is created and the
Finite-Element-Model (FEM) calculates where all rivets must be positioned to ensure rigidity of
the wing structure for the aircraft. This means that the hole pattern to be drilled by the robot(s)
are decided as well, hence, before even starting the operation planning, many robot tasks are
already known. The remaining operations include, choosing the cutter diameters and feed-rates
for drilling, reach-analysis of the robot(s), creation of via-tags to avoid collisions between robot
and surroundings. All these operations must be entered manually at present. It would be more
convenient to store all operation and process data in a database and apply a role based system to
let the system program itself. This would be described as the process-oriented abstraction level of
robot programming. Bolmsj (1989) define the tasks to be performed by the robot, in a task-
oriented programming language, as processes, where the programming language has enough

137
Introduction to Robot Programming and Simulation

knowledge to know how the processes need to be executed. In this dissertation task-oriented
programming is seen as equivalent to process-oriented programming. Nikoleris concluded (1991)
that robots are programmed in the object-level, even though the process programming is highly
desirable. The author of this dissertation concludes that object-oriented programming today is
emphasized on a large scale, but process-oriented programming has not yet had its breakthrough.

One way to employ process-oriented programming is by generating macros to define sub-process

tasks. Creating macros is essentially the same as creating a sequence of rules to be applied in
each stage of the operation planning stage. One of these rules could be:

Holes_dia() =import(wing_section_A_holes)

For each item in Hole_dia

If hole_dia == 10 Then
Cutter_dia = 10
ElseIf hole_dia == 12
Cutter_dia = 12
End if
Next

In the example above, an import function breaks down the attributes from a CAD geometry
called wing_section_A_holes. Each hole diameter on the CAD geometry is stored in an array
called Holes_dia. The hole attributes in Holes_dia is broken down to Cutter_dia. This can then
be ready to use for its purpose in the operation planning of the robot hole drilling when selecting
the right drill end-effector.

Concluding this section about the Process-oriented approach metaphorically, it could be said that
the robot knows about its tasks, whereby the robot is only given a start condition and an end
condition. Tasks to be performed by the robot are given by processes and not operations. The
robot knows by itself how to execute each process and manages to reach the goal by itself. This
enables the robot to work independently from a database where all attributes are stored. And
finally, to completely generate the robot paths, hence ensure no collisions with the surrounding
environment (i.e. production equipment and parts), the robot is made aware of the World Model.
Working with the Process-oriented abstraction level for robot programming, there are still several
pre-requisites required for the system to automatically create the operation planning.

Goal-oriented
The last of the abstraction levels of robot programming is the goal-oriented level. Today in the
development of operation planning systems, this level is far from being exploited. Presently, this
abstraction level is easier to interpolate if applied on autonomous robot systems, such as mobile
robot platforms or humanoids. Essentially, the robot is not only aware of the World Model but
also of the complete process at hand. The conditions from start to finish are not required, but
merely the start state and goal state. For instance, one instruction could be:

START: Go to building A and leave this letter on Alec Donssons desk

END: Report when back to base

138
 Introduction to Robot Programming and Simulation

The goal-oriented abstraction level requires abstract rules that are needed as prerequisites. From
an operation planning point of view for a manufacturing system, this would mean having
intelligent agents as part of the operation planning system, acting upon a knowledgebase filled
with process data. Russell and Norvig (2003) define an intelligent agent as:

An agent is anything that can be viewed as perceiving its environment through

sensors and acting upon that environment through actuators.

The sensors would be channels for the agent to perceive production prerequisites and data from a
CAD environment and the actuators would be functionality accessed in an operation planning
system that display 3D geometry to an operator for validation and corrections. A first generation
of intelligent operation planning should help the operator design the production cell, generate
collision free paths and suggest production data for optimal performance.
8.1.2 CATIA and DELMIA
For operation planning and robot programming in this research, the CAD system CATIA and
process planning system DELMIA have been used. Both CATIA and DELMIA are part of a
CAD/CAM products portfolio from Dassault System. Figure 8.3 shows the complete software
portfolio and the dashed rectangles show CATIA and DELMIA in the Dassault System
CAD/CAM portfolio.

(Enterprise Resource Planning) (Customer Relation Management)

(Supply Chain Management)
Figure 8.3: The Dassault System Product Portfolio for manufacturing simulation

DELMIAs Digital Manufacturing Solutions are built on the Dassault System Product, Process
and Resource (PPR) data model. This enables the continuous creation and validation of the
manufacturing process in the context of the product throughput and product lifecycle. The PPR
can be seen as a Manufacturing Hub, a pipeline that allows collaborative and concurrent
engineering early in the products lifecycle.

In this research CATIA and DELMIA are the two systems used for CAD design and
manufacturing simulation and process programming. DELMIA V5 Robotics is the main module

139
Introduction to Robot Programming and Simulation

used in this dissertation for robot simulation. One of the advantages using CATIA and DELMIA
together is the seamless integration between CAD and simulation. If something in the CAD
model needs modification during simulation, a simple double-click on the model brings the user
back to the CATIA modeling environment. In addition, due to the parameterized functionality in
CATIA, dimensions within the models can be updated by double-clicking on the entity that
brings up values on the dimension that can easily be changed. This is big improvement to using
unions for building CAD geometries in earlier generations of CAD systems.

140
 Operation Planning Methodologies

9 Operation Planning Methodologies

This chapter applying the introduction part of simulation and programming to this research.
First part presents the design methodology to building an Affordable Reconfigurable Tooling
system. Following this, the discussion on how the abstraction levels of programming correlate to
the programming method used in this research is discussed. The method of building the
kinematics in the virtual models of the Dynamic Modules is also presented.

9.1 Designing an ART system

Conventional tooling for airframe assembly is tailored to hold parts during assembly for one
product type only. The decisions made during the process of designing assembly tools require
years of experience. Today both the tooling designers and the product designers work in parallel
very early in the project. When the design of the tool starts, the assembly product is far from its
final design. The lead time required to design and manufacture an assembly tool is significant and
must be started well before the aircraft final product design is complete. Working in parallel has
been a typical methodology used within the aircraft industry for many years and is commonly
referred to as concurrent engineering.

The research of this dissertation has dedicated considerable effort to the development and
simplification of new reconfigurable assembly tools. The fundamental research questions include:

How can the CAD-system be structured to shorten time in deciding what tool components
are to be used?
What tools in the CAD system can be used to support evaluation of each step in the
tooling design process?
How can functionality be induced by building macros and functions inside the systems?
How can the operation planning platform support programming of, not only robots, but
also manual operations messages to operator and metrology functions?

The first stage of simplifying the development of new products was to initiate a different
methodology. The author, together with the ADFAST team, iteratively brought forth a design,
simulation and programming methodology. The methodology was divided into several steps. At
each step, meetings were held, to discuss alternative approaches. Figure 9.1 presents the step-by-
step methodology.

141
Operation Planning Methodologies

Existing assemblies in Step 1 Future assemblies in the

the same product Define Assembly Steps same product family
family
Step 2
Analyse Datums

Step 3
Analyse Gravity & Process Forces

Step 4
Robot Selection & Reach Analysis

Does MF need Yes Can existing No

rebuild? MF be used?
Yes

Step 5a Step 5a
Rebuild MF Select DM
from database

Step 5b Step 5b
Update selection of Select and build MF.
DM Mount DM on MF

Step 5c
Step 5 Select pick-ups from database or
Tool Design make new design

Step 6 Dynamic Simulation and FEM

Perform Dynamic simulation, reconfiguring simulation, airframe
assembly simulation, ergonomic simulation and FEM analysis

No Is the
tooling system
optimised?
Yes
Step 7
Figure 9.1: The step-by- Off-line programming
step methodology
142
 Operation Planning Methodologies

The step-by-step process is the culmination of engineering work that attempts to consider as
many factors as possible before deciding on a concept. This research used the presented process,
and after each step a meeting was addressed and a discussion was undertaken for choosing the
most appropriate concept. Further in this section each step is presented in more detail.
Step 1 Define Assembly Steps
Step 1 divides the assembly work into physical assembly stations and the assembly steps in each
station. In order to facilitate a smooth changeover to a new product type, it is important to
consider the existing production layout already in step 1. This can for example be difficult when
considering changing the production from that of a wing section to the production of a fuselage
section. However, it is also important to consider fitting a new product into the existing system
by delicately deciding how the assembly steps are determined. The output from step 1 is a
document of the assembly steps for each subsection.
Step 2 Analyze Datum Points
In the assembly of airframes, certain locations are more important to secure accuracy than others.
These decisions are analyzed in step 2, where the pick-up positions are being selected for all
the parts in each subsection. This step roughly considers airframe part weights and process forces
to logically balance the pick-up positions. At step 2, it is also important to consider tolerance
build up. Tolerance is considered both for pick-up positions and airframe part geometry in order
to achieve the dimensional requirements on the final assembly. The output from step 2 is CAD
geometries with datum frames positioned for each pick-up position and a list of tolerance
demands.
Step 3 Analyze Gravity and Process Forces
In step 3 the accumulated forces, in regards to both gravity and process forces, are thoroughly
analyzed. All the forces are transformed into total force vectors for each pick-up position. This
becomes the output from step 3.
Step 4 Analyze Robot Interactions
Another important input to the design of assembly tooling is to look at how robots are affecting
where to place Dynamic Modules and where not to place framework in order to maintain access.
Robot interactions consider both the reconfiguration of Dynamic Modules and the automated
processes, such as drilling. The most evident factor to consider is the ability for the robot to reach
a given location. Step 1-4 normally require many iterations to provide a solid foundation to
moving to the tooling design in Step 5.
Step 5 Tooling Design
Step 5 takes all the input from the earlier steps into consideration. The first question that needs to
be asked is; - Can the Modular Framework be reused from earlier projects? In a normal scenario,
a company using this technology would probably have an existing tool system already, and wants
to reuse it for their next generation of tools. Or, if the throughput is low, the company may want
to use one tool for several building steps for the same aircraft. The preferred scenario would be to
keep the exact configuration of the previously used framework. In practice it is likely that part of
the tool is rebuilt or at least the DMs need to be moved prior to configuration. It is not until step 5
that the tooling CAD geometries are brought into the CAD model. Further in step 5, there are two
alternative paths to choose from. The first path reuses the existing Modular Framework, while the
second path redesigns a new Modular Framework from scratch.

143
Operation Planning Methodologies

Modular Framework can be reused

If it is decided to keep the current framework you need to roughly analyze the prerequisites, such
as force vectors and robot reach, in relation to the already existing Modular Framework,. In any
case it would be preferable to keep the current configuration of the Modular Framework.

Step 5a. Rebuild

If the Modular Framework has to be changed geometrically to fit the prerequisites, then the
framework is rebuilt. Even if the framework does not have to be rebuilt, but the base plates of the
Dynamic Modules have to be moved, it is also considered a rebuild.

Step 5b. Update selection of Dynamic Modules

If no rebuild was necessary in step 5a, this proceeding step is only required to reconfigure the
existing DMs from the previous project. New DMs are used if the existing DMs do not manage to
facilitate all pick-up positions.
A new Modular Framework must be designed
If geometrically the changeover from one product to another is too great, such as in changing to
another product family, there may be a need to rebuild. In this case a new Modular Framework is
built using the concept of building from the aircraft outwards to the framework.

Step 5a. Select Dynamic Modules from Database

The Dynamic Modules are brought in from the ART database that was presented in Section 4.5.
At this stage only the force vectors and robot reach analysis need to be considered. The output
from step 5a provides an outline of how a Modular Framework should be designed to hold the
Dynamic Modules.

Step 5b. Select Framework Components from Database

From the output of step 5a, in step 5b, the Modular Framework is configured to hold the
Dynamic Modules in space. Step 5a and step 5b will require several iterations to result in an
optimal solution. It is important to understand that if no previous ART tool exists, there will be
greater freedom to select a good design.
5c. Selection pick-ups
As presented in Subsection 4.3.3, Pick-ups are allowed to be customized and vary for each
airframe they are allowed to be customized for each application. In the database presented in
Section 4.5 there exists many variants of Pick-ups, but it may be that there will be new Pick-ups
designed in step 5c. Experience from several case studies in this research showed that the Pick-
ups can be rather small and are relatively easy to design. It is important to understand that the
accuracy given by the robot is at the Capto Interface. Therefore, the distance between that
interface and the airframe must be as small as possible to avoid loss of tolerance. It is also
recommended to analyze tolerance build up with relation to how far the actual Pick-up position is
located away from the Capto Interface.

144
 Operation Planning Methodologies

Step 6 Dynamic Simulation, FEM and Ergonomic Simulation

Step 6 investigates the preliminary design of the tool in more detail. It is also an iterative
investigation of the preliminary design on the aspects of:

Dynamic simulation
Finite Element Analysis
Ergonomic aspects

So far in the ART step by step model, the tool is static. At this step, the first operation is to
transform the static tool to a mechanical device in the geometry simulation system. Using
CATIA, the tool is modeled using constraints. These are geometrical or dimensional relations
between two elements (parts and products). Using constraints is the method to define how parts
are related to each other in a CAD product. A feature built into CATIA, called Assembly
Constraints Conversion, transforms these constraints into joints of a mechanism. This is
performed by simply pressing one button. Now, the ART tool can be easily manipulated as an n-
axis machine in the simulation system. This makes it easy to change the tool configuration if the
robot does not reach, or the tool modules are obscuring for the part assembly analysis. It is in this
stage that the framework and DMs get their final locations.

Ergonomic analysis is also carried out at this stage, which affects the final design of the ART
tooling design. This dissertation is not covering ergonomic studies for the ART system. At this
stage, the methodology is covering the FEM testing of the framework and configuration of
Dynamic Modules. For a given framework configuration, the strength of the beams must be
analyzed. Either FEM tools built into CATIA, or that of another system can be used. The most
crucial item to investigate is the configuration of the Dynamic Modules. Current research is now
analyzing the strength of different Dynamic Modules in various configurations. The level of
analysis used for the demonstrator was restricted to a common sense approach. No complete
FEM analysis of the system setup has been undertaken. Individual physical tests of Dynamic
Modules are presented in (Simier and Rouyre, 2003; Amadori, 2003). The framework FEM
analysis is presented in (Pradeau, 2004).

Before creating the process-oriented program, presented in the proceeding Section 9.2, an in-
depth discussion is required if the tool is to be as optimized as possible. The final design of a tool
is not deterministic. The many alternative decisions that can be made during the design phase,
can change the outcome considerably if only one of the decisions is changed. Although iterations
within each of the steps are likely to happen, this step is where the development team must
discuss if the chosen tooling system will meet the customer requirement.

Step 7 Create the Process-oriented Program

At this stage the process-oriented programming is undertaken. Most of that work should already
be done from the Dynamic Simulations in the previous step. This step is basically putting it
together to one program that is downloaded to the workshop. The major work in this step should
be the process naming of the operations.

145
Operation Planning Methodologies

9.2 A Process-oriented Programming Approach

Programming robots is still a time-consuming process where the operator is still teaching the
robot, even if today the programmer uses a 3D-model of the Robot, as in the object-oriented
approach discussed in Section 8.1.1. In addition, the world model in a geometry simulation
system is a nominal world. The model always deviates from the existing real cell or becomes
different than the simulation model if manufactured. The inconstancy between virtual planning
and physical execution today is solved by calibration.

A reconfigurable manufacturing system is not exclusively about programming robots, but also
includes planning everything from manual operations, tool reconfigurations, metrology
measurement, etc. As the number of entities that need to be programmed and planned increases,
so does the complexity of operation planning. The author of this dissertation implemented a kind
off process-oriented programming approach to simplify the programming of robots and other
entities in the ART system.

Nikoleris (1991) makes a very interesting example of a vision towards how the graphical robot
programming environment should work. Nikoleris makes an example of how a robot welding
application should be implemented.

The programmer should be able to define the material type and the results of the welding
process and let the system derive the necessary process parameters such as current, voltage,
welding speed etc.

The approach to the research presented in this dissertation strived towards that vision in the
planning of operations for robots, manual work, measuring etc. In short, the approach to process-
oriented programming in this research constitutes the following user interactions in the graphical
robot programming system DELMIA.

1. The user has created an ART tooling design using constraints in the CAD system, which
is essentially step 6 in the ART methodology presented in Section 9.1.
2. The constraints are converted into the mechanism, where the static fixture is now possible
to manipulate as an n-axis machine, also presented in step 6 of the ART methodology.
3. The user brings in robots to the simulation environment, also considered in step 6 of the
ART methodology.
4. The Paths are defined by the user, and the tool can be easily adjusted according to point 2.
In the process of creating paths for the robot, via-tags are created. These via-tags are
named FlyBy (Further presented in Subsection 9.2.2).
5. The user defines operator messages by naming tags as OperatorMessage. Note that
these tags will not be processed by the robot in the execution of the workshop. Therefore,
the operator messages are created in the same location as the previous via-tag, or FlyBy
points (Further presented in Subsection 9.2.1).
6. The locations or tags that require high accuracy are named FlyIn, such as those
locations identified before docking to a Dynamic Module (Further presented in
Subsection 9.2.2).
7. User defines docking points to Dynamic Modules simply as DockingModule (Further
presented in Subsection 9.2.2).
8. User defines un-docking points as UndockModule (Further presented in Subsection
9.2.2).

146
 Operation Planning Methodologies

9. The tag, where the Dynamic Modules are reconfigured in a changeover, are named
ReconfigureModule (Further presented in Subsection 9.2.2).
10. In the initial location (the start position) of the robot there are as many tag points created
as there are Dynamic Modules. These tag points are named: ProbeModule, which will
instruct the operator running the integration software to measure all Dynamic Modules
using the 6D-Reflector. (Further presented in Subsection 9.2.1).

As a result of the steps above, a list of operations are defined. Before explaining the available
operations in the ART-process, Figure 9.2 is showing how the programming of the ART-
processes is performed.

Figure 9.2: The user interface for programming the ARTs manual and automated
operations

The user simulating, hence programming the ART system, simply double clicks on a tag frame in
the 3D-graphics. This will open a field box where the user enters the name of the operation to be
performed in that tag location. In the physical execution it does not matter where tag points for
the manual operations are located since the integration program handles what is to be performed
(i.e. robot movements or manual operations). For the purpose of simulating manual operations,
the tag frames are located where the user wants the robot to be located. This was a very
convenient way for the user to plan all the operations. The build in feature in DELMIA also
showed the user the trajectories for the robot and arrows for which direction the robot was
moving.

Next two subsections of these chapter presents the manual and automatic tasks in the process-
oriented programming approach.

147
Operation Planning Methodologies

9.2.1 Manual Tasks

Manual tasks are performed by an operator in the physical ART cell. Each operation name is
divided into three groups with each being separated by a colon. The first group is the operation
for the operator to perform; the second group defines which Dynamic Module to work with, and;
the third group can be a message to the operator or a tolerance requirement. The different kinds
of operations are presented below.

ProbeModule
Before the robot can start to configure or reconfigure the Dynamic Modules, the default location
of the Dynamic Modules must be measured. The operator attaches the 6D-Reflector on a
Dynamic Module and when the Reflector is attached on the Dynamic Module, the location is
measured. Figure 5.13c in Section 5.2.3 shows how the 6D-Reflector is attached to a Dynamic
Module. If the default configuration is outside the reach for the robot, the operator simply moves
the Dynamic Module into the reach envelope of the robot. A functionality built into the
integration program shows the user when the Dynamic Module is in the correct location. Section
6.4.3 shows how this operator menu works. This functionality was important so that the Dynamic
Module can be positioned as close to its nominal CAD location as possible to avoid unnecessary
singularities when the robot docks onto the module. Figure 9.3 shows the work envelope and the
red circle draws attention to the Dynamic Module outside the reach of the robot.

Figure 9.3: The Dynamic Module is outside the

work envelop of the robot

In this situation, the user simply reconfigures the Dynamic Module to be inside the yellow work
envelop sphere. This location is further transferred to the integration software the Probe Menu
(presented in Section 6.4.3) helps the operator prepare a good starting position of each Dynamic
Module. This feature was partly why the 6D-Reflector was designed with handles, such that the
operator can hold during the probing operation. Stored within the tag point, is the tolerance for
how close the operator should manually locate the Dynamic Module. The tag point naming
notation was:

ProbeModule:DMx:20:

The first group of the notation is the task name (ProbeModule). The second group is the name of
the Dynamic Module (i.e., DM1, DM2,., DM8). The last group, in millimeters (the value 20 in

148
 Operation Planning Methodologies

this example), was the tolerance in millimeters of how accurate the operator must locate the
module by hand.

OperatorMessage
Sometimes the user only wants to forward a message to the operator. There can be numerous
reasons for this. An example is:

Use Capto Chuck tool nr #2458372 when attaching the Pick-up,

or
Attach aircraft article Leading Edge #234522.

Before displaying an operator message, the robot is automatically stopped by the integration
program. As a result, the Operator Message is stored in the tag point prior this message in the list
of operations (to the left in Figure 9.2). The naming of an operator message tag is for example:

OperatorMessage::Attach aircraft article Leading Edge #234522:

The first group is the task (i.e. OperatorMessage), the second group is left empty, and the third
group is the actual message (i.e. Attach aircraft article Leading Edge #234522). For simulation
purposes, the tag frame is located where the user wants the robot located when showing the
operator message. Note that the DELMIA user does not see the message as a message box, but
only as part of the sequence of operations.

ApplyPickup
This task is the manual operation of attaching a Pick-up to the Dynamic Module (presented in
Section 4.3.3). This task is performed before moving aircraft parts into the tool cell. The task is
named:

ApplyPickup:DMx:Message:

The first group is the task name (i.e. ApplyPickup), the second group is the name of the Dynamic
Module at hand (i.e., DM1, DM2,.DM8), and the third group is optional, but can be used as
information to the operator.

9.2.2 Automatic Tasks

Automatic tasks are performed by the robot. The groups are structured the same way as the
manual tasks, where the user names the tag points for the operation to be performed. Each
operation is presented in detail below.

Docking
Logical to its name, the Docking task is performed when the robot is moving in towards a
Dynamic Module for the attachment of the male Capto interface of the Dynamic Module to the
robots automatic chuck. Normally this operation is performed prior to reconfiguration of the
Dynamic Modules, but it could also be used for attaching any end-effector to the robot chuck
(e.g. a drilling machine). The task naming is:

Docking:DMx:message:

149
Operation Planning Methodologies

The first group will make the integration software execute the case clause for performing the
docking sequence (i.e. Docking). The second group is to which module? (i.e., DM1,
DM2,.,DM8), and the third group is optional, but it could contain an operator message. Note
that the docking tag frame is updated from the ProbeModule operation after probing has been
performed by the operator.

UnDocking
The tag name UnDocking is basically the opposite to Docking. Hence, whatever is attached to the
robot chuck will be released. In the integration program, this operation will move the robot in a
negative TCP z-direction to make sure the robot is clearly moved away from the Capto interface.
The task naming is:

UnDocking:DMx:message:

In the current version of the integration program, the robot is instructed to use something called
SoftServo (for ABB robots), which essentially makes the robot flexible so that the Dynamic
Module is not forced in any direction during the UnDocking operation. The amount of softness
provided by the robot could be dynamic variable. For future versions, the level of softness could
be chosen by the user in the operation planning by placing an indentifier in group 2 or group 3 of
the naming tag.

FlyBy
The tag name FlyBy is moving the robot to a frame without measuring the location with the
metrology system. This is used when the accuracy of a location is not critical. In the integration
software this tag frame is not transformed to the tracker coordinate system {tr} because the
tracker does not need to perform any calculation. The tag naming is:

FlyBy::message:

The first group of the tag frame (i.e. FlyBy) will make the integration program move the robot to
the coordinate specified in the tag frame. The second group is empty and the third group is an
optional operator message.

FlyIn
The tag name FlyIn uses the metrology system to guide the robot to the correct location. This tag
frame is transformed into the metrology coordinate system enabling the metrology device to
calculate the deviation from the nominal value. The tag name is:

FlyIn:Accuracy:message:

The first group of the tag frame (FlyIn) will cause the integration program to move the robot to
the coordinate specified in the tag frame. The second group (i.e. Accuracy) is dynamic and
specifies the accuracy required for that coordinate frame. The integration program will not stop
the iteration of the control until the robot is within the specified value. The accuracy value
corresponds to the error: x 2 + y 2 + z 2 , hence the square root sum of the error. Section 6.5 is
explaining more about this accuracy value. The third group of the tag name is an optional

150
 Operation Planning Methodologies

operator message. This displays the integration program during the movement of the particular
FlyIn frame being executed.

ReconfigModule
The last tag name executed in the integration program is ReconfigModule. This tag name is
transformed to the metrology coordinate system, similar to FlyIn. Apart from moving the robot to
the frame location with metrology guidance, it also takes dynamically in accordance which
Dynamic Module to reconfigure. The tag name is:

ReconfigModule:DMx:message:

The first group of the tag name (i.e. ReconfigModule) will make the integration program move
the robot to the specified coordinate. The accuracy was defined statically during the current
version. The second group of the tag name (i.e., DM1, DM2,.,DM8) defines which Dynamic
Module is being reconfigured. Its specification is important for the integration program to enter
the correct subroutine, where certain locking sleeves are manipulated depending on which
module is being reconfigured. The third group is an optional operator message, which is shown
during the reconfiguration process.

9.2.3 The Transition between Virtual and Physical World

As the operation planning process is finished, it is time to move from the virtual world to the
physical world. This normally requires calibrating the virtual world model so that it matches the
physical world, or to calibrate the physical world to fit the virtual world. In the ART concept this
concern is rather trivial. This is because the robot is guided by the metrology system, where the
feedback-loop between tracker and robot will not stop until the required absolute accuracy is
provided. This can be compared to real-time calibration, where calibration is entwined into the
Metrology-integrated Control presented in Chapter 6. But what about the Modular Framework? It
was decided that the framework need only be positioned within a tolerance of centimeters since
the Dynamic Modules attached to the Modular Framework, allow a rather relaxed positioning of
the base plates. Because all Dynamic Modules are flexible in six degrees of freedom, the relative
location of all Capto interfaces positioned by the robot in the configuration/reconfiguration
operations, are within 0.05 mm if that is the required accuracy set in the ReconfigModule frames.
Essentially, the only calibration required is to inform the integration program where the the
Dynamic Modules are initially located. Recall that this was done by the probing operation of the
male Capto interfaces using the 6D-Reflector. From that point the integration program updates
the following tag frames:

The FlyBy frame just before the Docking

operation
The FlyIn frame just before the Docking
operation
The Docking frame

Figure 9.4, shows the sequence of operation list, where Figure 9.4: The tag frames in
the underlined rows show the frames that are updated. the operation list that are
The reason the FlyBy frame is updated is because it is updated from the Probing
located close to the docking location.

151
Operation Planning Methodologies

As a result of the operation planning in DELMIA, an XML file is created which contains Tag
frame names (i.e. FlyIn, FlyBy, Docking, ReconfigModule etc.), including the frame coordinates
in the robot base coordinate system and the forward kinematic robot joint values for each tag
frame. This XML-file is the only file required to be downloaded to the integration program. This
is because no brand specific robot program (i.e. ABBs RAPID language) is generated in the
operation planning or in the integration program, which was possible by the generic robot
programming approach presented in Subsection 6.1.2.

9.2.4 Final remarks on the process-oriented programming approach

A really process-oriented abstraction level of programming should generate the sequence of
operations autonomously, whereas the approach presented in the previous subsection is more an
objective-oriented approach using that context. Admittedly, the presented approach does not
interpret CAD models to automatically generate collision free robot paths that would be expected
in a truly process-oriented programming approach. Why the author has chosen to call this
approach process-oriented, is because the integration program interprets all processes defined in
the OLP system. Normally an OLP system generates brand specific robot programs that are
downloaded to a robot controller. In the presented approach there is no robot specific program
generated since the ART system is more than just about executing robot processes. The author
chose to use a more generic approach using the XML code generated in the background of
DELMIA so that it could be downloaded to the integration program including all ART processes.
This method shortened lead time in changes between the operation planning and the physical
execution where the complete ART process is defined at only one location by the user. From that
point, no additional work is required for executing the complete ART process to the high absolute
accuracies presented in Section 6.5. The author agrees this is not a pure process-oriented
programming approach as defined in Section 8.1.1, but it is striving towards that goal. The author
has performed research on auto-generated robotic drilling sequences presented in the sub-project
Gthberg and Stark (2002) supervised by the author. That research took frames for holes to be
drilled, defined in a layer in a ProEngineer CAD model, and auto-generated robot trajectories
using a program module called Rose. This was carried out in the OLP system RobCad to
automatically create via frames and hole drilling cycles. That research in the subproject went all
the way from the CAD model to physical execution within the robot, where the auto-generated
robot paths were downloaded as brand specific program to the robot controller. The RobCad
programming approach was not fully converted to the ART system using DELMIA, as presented
in this dissertation. This in part due to time limitations, but could in the future be incorporated in
ART system.

9.3 Definition of Kinematics in Dynamic Simulation

To accommodate dynamic simulation presented in step 6 Section 9.1, the Dynamic Modules had
to be moved in the simulation. Simulating a Dynamic Module means that the base plate of the
module is static and the Capto Interface is moving in the Cartesian space. To make this work the
Dynamic Modules must have the kinematic defined. Normally this can be supported by the
simulation software, but the DELMIA version V5R12, used for the simulations, did not support
inverse kinematics. Early in this research the modules were manipulated joint by joint. This
became cumbersome when placed in a complete ART system where DMs needed to be
positioned back and fourth many times. Furthermore, to analyze collisions in the reconfiguration
process, the Dynamic Modules were docked to the Robot Chuck in the simulation. For DELMIA,
the DMs are in fact considered as robots. Simulating reconfiguration means that two robots are
moved on the same trajectory simultaneously.

152
 Operation Planning Methodologies

DELMIA supported only forward kinematics (for homemade mechanisms). In the parallel
mechanical DMs this means that if one leg is moved all other legs in a parallel kinematic
mechanism also move. The author decided to develop a way to create inverse kinematics. The
Hexapod Dynamic Module will be used in this subsection as en example for explaining how the
inverse kinematics was implemented. The preliminary result for this methodology was first
presented in the subproject Wulf (2004) supervised by the author.

The V5R12 of DELMIA did support the possibility to close a kinematic chain. This can occur
when defining the kinematics of a parallel kinematic mechanism, such as the Hexapod. However,
in V5R12 of DELMIA, it is possible to manipulate, according to forward kinematics, one leg
called the master leg. The rest of the joints in the other five legs of the Hexapod are following as
slave joints. The inverse kinematics were created by adding six extra dummy joints on top of the
Hexapod (x, y, z, roll, pitch, yaw), which provide six additional degrees of freedom, as seen in
Figure 9.5.

Figure 9.5: Hexapod with a TCP coordinate system located at the top plate

The forward kinematics for the master leg of the Hexapod is further called the slave mechanism,
since it follows the movement of the dummy joints. The additional joints on the top plate were
the x, y and z axis, together forming a TCP. From this point, manipulation in x, y, z, rx, ry and rz
were possible. The three extra axes providing three prismatic and three revolute movements were
called fake robot, where the additional axes can be seen as a robot within the device. Moving

153
Operation Planning Methodologies

the fake robot in its joint axes, the Hexapod followed as a slave mechanism. Figure 9.5 shows
how the additional kinematic chain was configured.

The kinematic chain in Figure 9.5 started with three translational joints followed by three
rotational joints. To assure that the slave joints were following accordingly, the movements of the
master joints, the first new part (XY-translation) was connected via a translational joint to the
base plate. The last joint (rotational joint in z) was attached to a line located in the part of the top
plate itself.

This solution was applicable to all dynamic modules and made it easy for the system designer to
implement kinematics into devices whenever necessary.

There are limitations to this approach, which are related to the principle of how DELMIA handle
kinematic chains. All the joints of the kinematic chain behave in a correct way as long as the
values of the different joints are manipulated in the same order they have been defined during the
previous building phase of the chain itself. The first created joint in a chain always has the lowest
rank while the last created joint has the highest rank. This is important to know when jogging the
Hexapod on the linear axis. The current solution works as long as the joints with lower rank in
the kinematic chain are being manipulated first. If a joint with higher rank is manipulated prior a
joint with lower rank, the joints with lower rank will not move and the coordinate axes will slide
apart.

Simulating reconfiguration was made possible by synchronizing the movement with time steps.
V5R12 of DELMIA did not support two robots moving on the same trajectory. But two robots
could be moved simultaneously. This meant that the Hexapod robot and the industrial robot
appear to be moving attached in the simulation, since they both move to frames at the exact same
time step.

154
PART V: DICUSSION AND CONCLUSION

PART V
DISCUSSION AND CONCLUSION

Part V, starts with Chapter 10, the discussion of the possible improvement to
transform this research to a real industrialized assembly system. Chapter 11
discusses and answers the research questions in this dissertation. Chapter 11 also
discusses the overall conclusions of this dissertation and ends with discussion on
potential areas for continued research.

155
156
 Future Improvements

10 Future Improvements

This chapter discusses the remaining actions required to transform the technology that has been
developed in this research to an industrialized assembly system of a Company.

10.1 Introduction
What is the step required to move from a research lab facility to real production on a workshop
floor? Well, it can be anything from potential technical risk; limitations in the research platform;
design flaws; lack of quality etc. The authors opinion is that the role of the researcher is not to
provide turn-key solutions but rather a working model. This does not mean that there is any lack
of confidence in the systems ability to perform in a real-life production environment. This
research has not attempted to perfect existing technology solutions, but to innovatively combine
existing technology for completely new purposes. This means taking great technical risks which
should be taken within the confines a research lab and not on a production floor. It is like skiing:
if you never fall, you are not riding on the edge. In this research some part of what has been
developed are ready for the transition from the lab floor to the production floor, but there are also
some areas that need further research and development before leaving the lab with the fork lift.
This chapter is going to identify what improvements are needed before final implementation can
be realized.

10.2 Modular Framework

The Modular Framework (MF), uses beams as a surrounding system to the Dynamic Modules,
and because of its robustness, has proven to work well in the lab environment so far. Pradeau
(2004) discussed the problem with the low friction coefficient in the protective coating used to
protect the beams from corrosion. Pradeau states that other kinds of coatings can be a good
compromise between protection and adhesion. A Zinc coating is one alternative, since it has a
better friction coefficient than steel in such case.

The dimensions of the beams were deliberately chosen to have an over constrained size in order
to make sure the framework did not affect the movement of the Dynamic Modules.
It is very important that the Modular Framework is very stiff if the pick-up of the Dynamic
Module is moved because of load. Otherwise it becomes complex to distinguish if the movement
is a function of the Modular Framework or of the Dynamic Module itself. Further research is
required for the analysis of the framework stiffness properties so that the framework final design
can be evaluated by a set of simple equations. These studies will provide exact stiffness
properties for each Box-joint and clear instructions on how much torque each screw in the Box-
joint can be tightened to. From the conclusions of this research so far, the weak spot is not the
strength of the beams, but in the strength of the Box-joints.

The demonstrator presented in Chapter 4 was supported by air-cushions for the transport of the
Dynamic Framework on the floor. In an industrial implementation the design should consider
moving the fixture or the robot. Preliminary tests have been undertaken in the robot lab at

157
Future Improvements

Linkping University, moving a robot to enlarge its work volume. It is believed that this area will
require further study to conclude the optimal solution and will be driven by a particular
production scenario.

10.3 Dynamic Modules

Dynamic Modules described in this dissertation have a number of areas where improvements can
be made prior to industrialization. This section will bring up some of the most interesting areas.

10.3.1 Ball-joints
The Ball-joint sub-modules did show movement as the locking sleeves were activated. This
resulted from the kind of locking concept used. When the locking sleeve within the ball-joint is
activated, a piston is moved by hydraulic pressure and pushes the ball onto the upper casing. Bear
in mind, that prior to locking, the robot has positioned the Capto interface of the ball to high
accuracy. This means that locking must not affect that accuracy. Experiments showed that the
ball joint moved in the range of 0.3 mm. This measure did vary slightly if the robot had dragged
or pushed the ball prior to locking. There are many ways to overcome this dilemma. One method
tested was to measure the location of the ball joint before and after locking whilst the robot was
docked to the module. The movement was registered and the locking deactivated and re-adjusted
according to the negative locking movement and locked again. In this case the locking movement
moved the ball closer to the nominal value. Methods like these need to be thoroughly investigated
before industrial implementation. Preferably, the ball-joint locking design should be modified to
avoid any movement during locking.

10.3.2 Un-docking a DM
When a DM is configured or re-configured by the robot, un-docking is carried out to release the
robot chuck from the Capto interface. The design of the robot chuck causes a release movement
of 0.3 mm. As the robot chuck is locking, this movement is necessary to push the Capto interface
hard against the zero plane of the female and male Capto and thus provides 2 m repetitive
accuracy. In releasing the pressure, the 0.3mm creates a pushing force on the already configured
DM. This research assumes that the robot is weaker than the DM such that the 0.3 mm movement
is absorbed by the robot. However, some un-expected movement can occur of the DMs during
undocking. In an industrially designed installation, the DMs would not move on this assumption,
but nevertheless, it is important to be aware of the alternative scenario. With the implementation
of the robot, a function called softservo softens the manipulator on demand and in effect dampens
the force of the pulse.

Un-docking is the process of moving the robot away from the Capto interface. In the same
context as the locking and release pulse, the robot must not collide with the Capto interface while
retrieving it in an un-docking sequence. Using an ABB robot, the command RelTool is useful for
reducing the risk of collision when the robot is moved in the negative z-axis via tool coordinates.

10.3.3 Bearings
To maintain the high accuracy in a reconfigured DM, it is necessary to have sufficient rigidity in
the Module. The rigidity of a DM is governed by the bearing size and its ability to resist the load
exerted on it. Early in the project, some modules had rigidity problems due to incorrect bearing
selection. Some of the DMs where not designed to resist large radial forces. This was especially

158
 Future Improvements

the case for serial linked DMs, which have a larger tolerance interval for radial movement as
compared to the parallel mechanical type.

10.4 Pick-ups
In Chapter 4 was discussed that the male Capto interface on top of the Dynamic Modules cannot
themselves be used to hold the aircraft parts. Instead a pick-up is used to hold the aircraft parts,
which are attached to the male Capto interface on the Dynamic Module using a female Capto
interface built in to the pick-up. Since it is the male Capto interface of the Dynamic Module that
is positioned with high accuracy by the robot, the pick-up has to be calibrated in a CMM, where
the relative distance between the female Capto interface of the pick-up and the holes (used for
attaching the aircraft parts) are measured.

In the demonstrator presented in Section 4.4 this CMM calibration worked fine, but there is a
potential risk to make mistakes in this operation. If this operation is done wrong, the absolute
accuracy of the pick-ups locations, and thus the quality of the airframe assembly will be
jeopardized.

10.5 System Integration

When synchronizing the robot and the metrology system, different transformation matrixes are
used. With its orientation (frame), a transformation matrix converts a point from one coordinate
system to another. During the transformation between metrology- and robot base coordinates,
both the metrology system and robot are working in the same coordinate frame. The
transformation matrixes were created by calibrating the robot and the tracker using several
measurements of the robot TCP locations throughout its work envelope. Due to the relatively
poor absolute accuracy of the robot, errors in the transformation matrix occur between it and the
metrology system. This error varies depending on where the robot is working in the envelope. It
can vary between ~0 and 1.5 mm.

One way of reducing this error is to use a robot which provides a better absolute accuracy.
Alternatively, it is possible to update the robot calibration by making extensive calibrations to the
robot calibration parameters. The latter is about generating a good static model of the robot. The
advantage of having a metrology system online with the robot is that a bad static robot model
only affects the time taken to iterate its position to a high accuracy point. Since all calculations
carried out in the tracker coordinate system, the error is eliminated by the controller. The tracker
simply continues to reiterate until the robot is located within the tolerance requirements.
However, in an industrial implementation, a good static model may be preferred to reduce cycle
times.

10.6 Operation Planning

The ART system is programmed using the offline programming package DELMIA. The only
calibration required, after changing from a current program to a new program, is through the use
of the 6D-Reflector probe. This means that the integration software runs on nominal data with the
trust that the metrology system positions the robot and parts within a specified tolerance on a
nominal frame. This means that the digital model in the simulation package must be accurate. It
may appear trivial, but there is a risk to make mistakes by creating the tags within the simulation
model. If one tag is misplaced by just a few micrometers, or rotated slightly wrong, the accuracy

159
Future Improvements

of the system will be affected. Some parts in the digital model of the system are less important
than others.

The process-programming abstraction level needs to be fully implemented to make programming

of a drilling sequence faster in an industrial implementation. Only preliminary research in auto-
generated collision free robot trajectories has been done by the author. This concept never was
implemented in the ART system, due to lack of time. To have a really flexible and time efficient
programming concept, this method must be implemented in a final Affordable Automation
solution. Teaching the robot in 3-Dimensions is time consuming if an airframe product drilling
sequence is to be programmed. A typical aileron structure has more than 1000 holes that need to
be drilled. For the programmer, that transpires to a significant amount of repetitive mouse
clicking. Presently, the programming approach presented in Chapter 9, is effective in that there is
no calibration required other than the DM probing, For this effectiveness to be realized, the
programming of drill sequences must be fully autonomous.

160
 Discussion and Conclusions

11 Discussion and Conclusions

This chapter begins with answering the research questions in this dissertation followed with a
discussion of the overall conclusions of this research. The final section of this chapter discusses
potential areas for continued research.

11.1 Critical Review of the Research Questions, Objectives and Aim

In this section, the discussion focuses on how this dissertation has addressed the research
questions, objectives and aim posed in Section 1.2. Based on this discussion, the reader may
critically review progress made towards answering these questions and fulfilling these goals.

1. How can new drilling technology enable the use of industrial robots for aircraft
assembly automation?

From what was concluded already at the end of Chapter 2, off-the shelf industrial robots cannot
cope with forces other than gravitational load; hence, to guarantee maintained high accuracy, low
forces in the robot processes are crucial. Chapter 3 presented the first of the key enabling
technologies in this dissertation the Orbital Drilling method. In Orbital Drilling, the cutter
rotates in an orbital movement around the hole center. As the feeding movement commences, the
cutter is moved in a spiral movement through the stack of materials. Due to the high rpm of the
cutter, Orbital Drilling was explained in Chapter 3 as a high-speed machining process, which is
well known today to cause only small cutting forces. Chapter 3 concluded from force
measurement that Orbital Drilling has 8-10 times lower forces compared to Conventional
Drilling. Although forces are low in Orbital Drilling, the second part of Chapter 3 discussed the
need to use a pressure foot for securing the robot to the surface, although the forces are small.
Based on the authors experience and from discussion with engineering professionals in the
aerospace industry, the air gap between the materials in a stack of materials is required to be
removed in order to hinder burrs which can enter between the materials. This is achieved using a
pressure foot, just for the sake of eliminate the air gap, even if the cutting forces are zero. When
employing a pressure foot, the force will not be gravitational; hence, the robot cannot compensate
for this force and will cause a deflecting movement within the robot, essentially in six degrees of
freedom. The author recommends using either a force cone concept, presented in the final
remarks of Chapter 3, or a metrology system, as presented in Chapter 5, to detect deviation and
compensate accordingly.

In summary, the short answer to this first research question is that one solution is to use Orbital
Drilling. Having said this, the author is not excluding Conventional Drilling as a technology
suitable for automation. Nevertheless, from the perspective of this research low process forces are
an advantageous approach for industrial robot drilling automation. The fact that deburring can, in
some applications, be eliminated using Orbital drilling supports yet another perspective of this
research to reduce lead time and hence achieve low-cost, both integral parts of the main
objectives of this research.

161
Discussion and Conclusions

2. How can new assembly tooling be developed that uses an industrial robot to perform
changeover between products within a product family, and with a short changeover
time?

As presented in Chapter 4, a new tooling concept called ART was developed during the course of
this research. One part of the ART concept was described as comprised of Dynamic Modules
(DMs) that are flexible in six degrees of freedom, and flexible in the range of decimeters. The
movement in the DMs was performed by the robot. The robot docks and relocates the top end of
each DM. The middle of Chapter 4 explains how the Capto System, which is the tool changer
system used in this research, makes docking guarantee no loss of accuracy. Uncertainty in
accuracy may only be questioned from the robots positioning, not in the tool changing system.
This is important, since it is the robot that provides the accuracy in this context. The tool changer
used in this research the Capto System had a repetitive accuracy of 2 m, which is small in
comparison with the system absolute accuracy of +/- 50 m.

Chapter 4 presented three different parallel mechanical and four serial-linked DMs, all modules
flexible in six degrees of freedom. Chapter 10 presented some remaining shortcomings before
they are ready to be industrialized. There are some solutions that require further development due
to movement during locking and unlocking, stiffness issues, and play in the mechanism itself.
The solution that showed the most trustworthy properties was the Hexapod. The Octapod also
showed robust working conditions, but the downside to that solution were the stiffness problems
in the design itself. However, the Octapod was the first module to be built, and has provided
many lessons that were applied to the other DM. The serial-linked DMs worked well given the
right conditions. Certainly, a serial linked DM should not be exerted a load radically on its top
end point, which will cause deflection. By using DMs in the correct manners, the serial linked
DMs have a purpose to fulfill. These scenarios are access to complex structures, holding contour
boards, pick-up support with low forces, or higher forces if the force is pointing in the direction
(or opposite direction) to the DM. In the first part of Chapter 4, the bed-of-nail configuration that
used only 1-axis DMs was discussed.

The author would like to point out one advantage with modularity for a company that is
interested in implementing the ART solution. This research has focused on modularity, which
had its advantages in reuse, ability to upscale or downscale parts of the modules, or to change
parts of the modules between concepts. One example is the hybrid solution, i.e. between a serial-
linked module and parallel mechanic structure. Moreover, the modularity creates freedom for a
company to slowly implement this technology. Starting to use the ART solution does not mean
everything that is presented in this dissertation must be included. A typical company,
recommends the author, should start by bringing in one of the DM concepts into existing
production, partly for evaluation but also to give time to learn about a new technology, such as
the solution presented in this research.

When the author went to conferences in the early conceptual stage in the development of the
ART solution, there was some skepticism concerning the idea of using the robot to reconfigure a
fixture. Some said that the robot would not be able to be reconfigured due to the friction in the
module. That problem has not been discussed in this dissertation because it has not been a
problem. The robot used in this research has shown great ability to move in small increments.
Resolution in robots was briefly discussed at the end of Chapter 2. The author was concerned
about the opinions that the robot would not be able to make small movements under friction,

162
 Discussion and Conclusions

which is the case when being docked with a DM. But early experiments showed this to not be a
problem. This was further proven by moving the robot with a docked DM to a resolution far less
than the 50 m that was discussed in the final part of Chapter 6.

Furthermore, the Physical Demonstrator section in Chapter 4 describes how the hydraulic system
for locking and unlocking DMs was implemented. At the current state, there are different
hydraulic pressures for different modules, and even for pneumatic locking sleeves. This can be
explained from two perspectives. Firstly, in early stages in the project, the locking sleeves from
ETP Transmission AB had different locking- and locking pressures. As new DMs were
developed, ETP developed locking sleeves with the same locking and unlocking pressure. The
second part of the research was to evaluate different solutions for locking, which include
pneumatic locking sleeves.

The answer to the second research question in this dissertation is answered trough the DMs. The
authors declaration of what product family means was discussed in the beginning of Chapter
2. Typical product family examples include wings and body structures. The DMs can be changed
within the geometrical range that is pertaining to one product family. When the range between a
changeover within a product family is beyond the reach for one DM, redundancy of DMs is
applied, which means that for one particular building step, some DMs are not used.

3. How can the assembly tooling in research question two be developed to enable
changeover between product families?

The answers from research question two leave one question open for discussion. Changeovers
between product types within a product family are managed with the DMs, but what if the
company using the ART solution is making changes between product families? As presented in
Chapter 3, the framework surrounding the product and carrying the DMs are themselves modular,
and are hence given the name Modular Framework (MF). As presented in Chapter 3, the Box-
joint solution has two purposes; the first to eliminate welding for attaching the beams and the
second to attach the DMs on the MF. The first advantage, as presented in Chapter 4, was the
short lead time in building the actual tool.

The lead time of building the ART system was discussed briefly in Chapter 4, in the Physical
Demonstrator section. The author, together with four students, built the complete framework of
the physical demonstrator in one working day. This was done without using a roof lift, where the
vertical beams of the MF were equipped with a manual winch system, and the horizontal beams
were manually winched up to the proper location.

4. How can the absolute accuracy of industrial robots be increased to maintain the
demands placed on them by aircraft applications?

In Chapter 2, Orbital Drilling was introduced as a drilling method to drill high accuracy holes. In
Chapter 3, robot-manipulated tooling was introduced as a technology which
configures/reconfigures DMs to high accuracy. In Chapter 2, the problem with low accuracy in
industrial robots was discussed. The third key enabling technology the Metrology-integrated
Robot Control was introduced in Chapter 5 as a solution to the hampering issues with low
accuracy in industrial robots. Chapter 5 introduced existing metrology systems, and thus not just
the metrology system chosen for this research, which gave the reader the possibility to

163
Discussion and Conclusions

understand that there is more than one system alternative. Common to the different metrology
systems presented in the beginning of Chapter 5 is that they already exist on the aircraft
manufacturing floor for calibration use. This means that the Affordable Automation approach
presented in this dissertation does not have to burden the whole investment cost for a metrology
system. This is crucial, as metrology systems today are expensive; therefore, it is important to
utilize these systems 100%. In fact, the author has noticed that at a typical airframe-building
company today, its metrology system is not 100% utilized. One reason for this is the fact that the
metrology systems are needed more in the industrialization of new products, which is only under
certain periods. The rest of the time, the systems are only used for regular calibrations of fixtures.
There are also other examples where the metrology systems are used for other purposes, such as
final body join, where a metrology system is used to locate the aircraft parts relative to each
other.

The second part of Chapter 5 presents the metrology system used for this research the Leica
LTD800 that was presented to measure in 6 degrees of freedom (6DOF) by the use of an
additional camera on top of the laser tracker unit to measure the orientation of a probe or
reflector. Leica Geosystems from Switzerland introduced this camera system in the ADFAST
project, where the author had the honor to be the first to use the new system outside Switzerland.

The last part of Chapter 5 brought up a discussion on how the approach to Metrology-integrated
Robot Control relates to other concepts of metrology-integrated robotics, and which show that
there is similar research to that presented in this dissertation, but for slightly different purposes.
One example of similar research that was presented used 3D image metrology to guide a Tricept
robot for machining, while another example used laser metrology for industrial robots to fine
position a drill cutter with two extra degrees of freedom built in to the drilling end-effector.

In this research, maintained accuracy in robots differs from what is normally meant with high-
accuracy robots. The author brought up this discussion at the end of Chapter 1. The essence from
that discussion was basically that robots today are calibrated using metrology systems such the
laser tracker used in this research. The author postulated that that kind of calibration is, however,
static. The author continued further in the discussion that a statically calibrated robot presupposes
a static environment. Situations like the ones presented in Chapter 2 (drilling) and Chapter 3
(reconfigurable tooling) do not describe a static environment. The authors view is that todays
robots are not designed to cope with dynamic loads. Today robots are designed to carry hanging
(gravitational) loads attached to the robot, whereas contact scenarios, which means that the robot
is engaging a surface prior to drilling, or docking to a flexible fixture module, is not a hanging
load, but a dynamic load. To cope with the two latter scenarios, Chapter 5 introduced the
Metrology-integrated Robot Control.

Next, Chapter 6 explained how the robot and metrology system were integrated. A conventional
TCP/IP network was the infrastructure for the integration. Embedded off-the-shelf systems, such
as the emScon from Leica Geosystems and WebWare from ABB, were used to make a seamless
integration between the two systems. Chapter 6 also declared the different coordinate systems
and transform equation that constitute the mathematical platform for the calculations running in
the integration program that execute the system. Later in Chapter 6, integration software was
introduced and explained. The last part of Chapter 6 presented how the feedback loop between
the metrology system and robot was implemented.

164
 Discussion and Conclusions

In Chapter 7 the calibration chapter the author derived the different transform equations
necessary to perform calculations of the calibration procedures. It is easy to assume that all these
calibration procedures need to be performed every time the system is programmed from the
operation planning system. That is not the case. It is important to understand that between each
program, i.e. each reconfiguration, the only calibration procedure is the probing of Dynamic
Modules. All other calibration procedures are only done when the Affordable Automation system
is setup for the first time. But if a new end-effector is manufactured and so forth, it needs to be
calibrated to work together with the Affordable Automation system. In addition, if the tracker or
robot is moved, certainly the part concerning metrology system and robot requires recalibration.

From the discussion so far, the summary to answering the fourth research question is to
incorporate any of the already existing metrology systems, today used for fixture calibration, with
the robot controller, to guide the robot to high accuracy. This research integrated a Leica LTD800
with the robot controller, which has the capability to measure the robot continuously in 6 degrees
of freedom (x,y,z, roll, pitch, and jaw).

5. How can new operation planning methods simplify programming of the production
system presented in this research?

In Chapter 8, the simulation and robot programming system was presented. To assist the reader in
understanding the scope of the research in this area, Chapter 8 started out by introducing the
reader to robot programming by discussing the abstraction levels of robot programming. The last
part of Chapter 8 briefly introduced the platform being used in this research for operation
planning.

Chapter 9 continued by presenting the methodology for planning all operations in the Affordable
Automation approach, and not just the robot trajectory planning. This methodology reaches all
the way from initial studies of the airframe parts to be assembled, to a completely designed and
programmed ART system, including operator messages and instructions for manual operation,
robot operations and measurement operations for the tracker. A discussion was made whether this
approach is a process-oriented abstraction level, or purely an object-oriented abstraction level of
programming. The conclusion was made that the system in its current state is more of an
approach towards the process-oriented abstraction level than a pure method. Essentially, process-
oriented programming in the context of programming the Affordable Automation system would
require autonomous generation of collision-free robot trajectories, where decisions about if the
robot should be moved online with the tracker or not are made. Furthermore, the operator
messages and instructions would be automatically generated, where the system would
understand that manual operations need to be made. In this research so far, this intelligence
does not exist, and is more a kind of an object-oriented approach at current state. The reason the
process-oriented programming is postulated by the author as an approach in this dissertation; it is
because the structured process naming of operations is striving towards process-oriented
abstraction level of programming. Furthermore, taking the step to automate the programming of
all the processes and robot trajectories is not a big step, and would be a pure process-oriented
programming method. This matter is further discussed in the Section 11.3, Continued Research.

The fifth research question is answered in this research in four different ways. First, using a
structured method from the first stage in the tooling concept generation stage to the finally
programmed system supports the user in the decision-making process. Normally, making

165
Discussion and Conclusions

decisions to generate rough designs take years of training and experience. A methodology
shortens the learning curve in designing a rough concept of the tool. Secondly, in the process of
fine tuning the rough concept of the tool, by transforming the constraints in the CAD system to a
mechanism that is presented in Chapter 9, it becomes trivial for the user to modify the tool to a
final configuration. The tool in this stage is like a n-axis machine, were each axis can be
explicitly coordinated, either by dragging the beams directly in the 3D graphics, or by inserting
exact coordinates, where kinematics of the tool are defined. The last part of Chapter 9 explains
how the inverse kinematics were implemented to manipulate the DMs in the Cartesian coordinate
space. Thirdly, when the complete system is finished, including a configured Modular
Framework and Dynamic Modules, robot movement and decision on manual operation, the
system is programmed by naming tag frames directly in the 3D graphics. Fourthly, there is no
type-specific robot program generated at any time in the system. The generic robot programming
method presented at the beginning of Chapter 6 does not require the general robot movements in
the simulation program to be post-processed (transformed) to robot-specific program code, but
instead, as explained in Chapter 6, is executed in task routines in the integration program.
Instead, the universal XML-script that is stored in the background of the simulation is
downloaded from the simulation environment to the integration program. As was explained in
Chapter 9, this made the transition between the virtual world and the physical world purely
seamless.

11.2 Discussion of the Research Results

The approach in this research to Affordable Automation has proven, in virtual simulations, to
also work for other vehicles besides aircraft. The subproject discussed in Section 4.6 showed the
technical success of the ART system; the only other matter is if the scenario is suitable for it.
Without any major efforts in analyzing the automotive field for applicability, there were viewed
to be three production scenarios that would benefit from using ART technology. The first one is
when building prototype cars; the second one concept cars; and the third one control-
measurement fixtures. The ART system is not only fast when it comes to planning processes, but
also when building the products themselves. The most evident advantage would be to reconfigure
the tool to build several sections of the car in one fixture. The author concludes that the
affordable automation approach is easy to transfer to other vehicle fields. However, it is
important to analyze the throughput for the particular scenario. Reconfiguring an ART system
takes approximately one hour, which would indicate that the ART system is not likely to be
implemented directly in the serial production of mass-produced cars, but instead for the three
scenarios discussed above.

How is the industry currently meeting this approach to Affordable Automation? Current research
in the projects involving the author that is not discussed in this dissertation concerns in part
participating in the industrialization of the ART concept to the next generation of aircraft at Saab
Aerostructures. Due to confidential agreements, the author cannot expand this discussion in
detail, the fact is that the ART system will stand on the production floor at Saab Aerostructures
within a four year period. The first generation of the ART solution will be somewhat simplified,
i.e. not having a robot to configure/reconfigure the Dynamic Modules.

Besides those at Saab Aerostructures, the author has met people that think ART is an interesting
idea. The Stork Fokker Company in the Netherlands found the authors paper Affordable
Reconfigurable Tooling (Paper III) interesting. Today, Stork Fokker has a KUKA robot
configuring/reconfiguring a one-axis bed of nail tooling solution, which has similarities to the

166
 Discussion and Conclusions

ART system. Instead of using the hydro-mechanical locking sleeves as in the ART system, Stork
Fokker uses pneumatics and manually-operated handles for locking the bed-of-nail sticks. They
are currently reconfiguring the tool to seven different components: outer flap, inner flap and
aileron, all of these right and left hand side of the aircraft. The seventh component is a rudder
section (Muys, 2005).

Furthermore, Airbus UK, which was a partner in the ADFAST project, has shown great interest
in guiding robots with a metrology system. At Airbus UK, it is not a question if, but when serial
kinematic robots are moving into production. A solution, such as the one presented by the author
having a laser tracker to guide the robot, may seam somewhat expensive. But for the aerospace
industry, the trackers are already on the workshop floor performing fixture calibration; hence,
using a tracker to guide the robot would only increase the utility of the metrology system.

11.3 Continuing Research

Section 6.5 presented limitations in the synchronization speed between the metrology system and
the robot controller. A feedback loop of 3-5 seconds, as of today, enables robots to reach and
maintain high accuracy for Affordable Automation. The logical next step in this research is to
increase the speed of the feedback loop. This will place new demands on both the metrology
system and the robot controller. Current research by the author (but not part of this
dissertation)and the research team at Linkping University has made a first step towards
increasing the access to the robot controller. Preliminary results on this topic can be found in
Sreling and Lundqvist (2005). This is an implementation being developed at the robot research
group at Lund University. The joint research is part of the FlexAA project shown as future
research projects involving the author. The conclusion of that research so far is that it is possible
to access the robot controller, in real-time, with a feedback loop of 250 Hz, which is a great step
forward compared with todays 10 Hz non-real-time system in off-the-shell industrial robots.
The high feedback loop speed will open up many new fields for exploring the potential of using
industrial robots for aerospace manufacturing applications. Furthermore, this will make the
metrology system the bottleneck in the area of Metrology-integrated robot control. Briefly
mentioned in paper IV, a system like Leica does provide possibilities, in future versions, to
access data more rapidly than todays 4 packages with data per second that is delivered through
the emScon server in the tracker controller. One solution that was discussed in paper IV was to
use a new interface directly to the tracker controller. In this setup, the emScon server is only used
to initialize and control the measurement process, but data would be streamed in high bandwidth
directly from the tracker controller to the master computer. These are indeed interesting areas for
future research in this field.

Chapter 10 discussed some of the shortcomings of the ART system. The robot research group at
Linkping University has several paths to walk from this stage. One interesting path is to
continue Saab Aerostructures industrialization of the ART system, but this is a road with many
challenges ahead. Saab Aerostructures selection of the Hexapod certainly places a lot of focus on
its capability. But we as a university have a responsibility to continue further research in the area
of improving the capability of the serial-linked Dynamic Modules in order to bring forth the
complete concept of Affordable Automation. In that process, there are many new industrial
segments that have shown interest in this type of flexible tooling. The automotive industry has
shown interest, but the author thinks that the concept of Affordable Automation can be applicable
to many areas of vehicle assembly, given a suitable scenario. An approach for continued research

167
Discussion and Conclusions

could be to investigate ART, similar to what was done in the VOLVO case presented in Chapter
4, but also for other Vehicle assembly areas as well.

The discussion in Chapter 9 concluded that the current version of the Operation Planning system
is more a kind of object-oriented abstraction level of programming than pure robot-oriented;
hence, the author chose to call it an approach towards process-oriented programming. Separate
projects from the Affordable Automation concept have studied these areas, making smaller
research attempts. The author sees great potential to continue improving and exploiting this area
for new findings. Future research will focus on bringing in parametrisized CAD models that carry
production related parameters, such as hole definitions, hole quality requirements, datum
accuracy etc. The simulation package shall take these attributes, and by intelligence automatically
generate robot trajectories, manual operations, operator instructions, drilling parameters etc.
using a so-called intelligent agent (commonly used in Artificial Intelligence) between the CAD
and CAM systems. Part of this research will be driven forward in the Finabox project that was
presented as a future project in Section 1.4.2.

168
 References

12 References

! "#$ % &

' (( (
)*"#$ (# + (# + (&

' ,(
),"#$ - &

.. ( / ( 0 ! " # $% &
0 1 * (2!2 - &

* 3 45 -46 + 5 5 % 5 * *
4 -7
// & & / /-45 -46& - &

85 5 9 : ; ( +55 % 5 5 * *
4 -7
// & & / /< & - &

5 9& * = + + 5-4 4 5 7 5 9& 5 >. :

< <0 < 4= 5 - 5 *&

?& ) ! ' (
9 .4 0 .@ '? "1 A,&

( ' & 0 = - 6B # + (
4 -7
// & & &

@& ; 5 & ,2* ) ) '

< 6 - 9" B 59 <
9 4 ;5& -- * *!

: : & *' + ,+ 45
.4 0 .@ '? '#8 "C /( " D 6

15 & % 0 < < 5

+ = ;5 4-
- - E < 4 )) ' 56 - % !
&

6 @& & ,,, - & " -

; <= > "1- &

169
References

F & ,,( "." ! . / !" 0 .4 # & (A*

0 :G- <H + 6&

<= 5 -5 8 - * B -< + * ( )
4 -7
// & <& / 5
/ < I / $ !/ &4 5&

<3 = 9 : % 5 : ( +55 5 5 * *
& <& / 5/ &

5 F
G 8& & ,2, " " ' # , (( 2* A&

D& J4 < D& ,,, 0) ! ' 1 2

D 5 <5 9 < 6 / -& ) )A&

: <4 & ) , ", )(4 ' 56 -

' D =4 < H &

: <4 & 8 4 & ) ", , 3 ' E

5 = <K - 9 5L = &

5
5 8&#& ,,! 4 $ +
'+ 69 < D 5
M; 5
5 =4 4L &

5
5 #& ,,, ' - 9 <
. 4 5<6 = > "1- D ,,, 5
5+ B H 7
55 <> ; <&

56 8& 5
-4 %& # < & B5
5 & = D& ,,, 5 '
, 0 - 9 <. 4 5<6= >
"1- D &

65 9& < + - G 5 D ), 8 - N .4 7
0 .@ '? '#8 "1O / O "&

= : &'& ,,A ) & '""" 9 <I

) 7( )

= < D&D& ,2, B 56 ' #

,* 2 ,&

= 4 &D& # & 5 & 0 8& 8 & & # 6 D&

( 6 6 & ! ' " - ; <
= > "1- - ) &0 H &

170
 References

= 4 &D& * = + < + 7= 4
& @ : @ +55 D 69 5 5&

=H9H0' ' 7 ' N; 5 5 - 4

F 7" - 0. , ( -& &

P< 5 < "& 4I & ;& & ,,(

--5 + < &

5 "& (& 7 4 & 8 9 " ( -

9 <> ; <= > "14 - 0
9% H &

5< & ?& ; : "& ' D& *: 0 ;*: < 0 =

" - ; <= > "1-
% )&

" 6 5- % 5 * % 5 6 + -5 *
4 -7
// 4& & / / 5Q ' R, * )*2>S 6R 5<6> R &

" < G 9 ,,2 > ) ? " ' # +5

5 + T S
T S #& ,,2 ) 6 = <&

" : '& * = + 5-4 4" : 7" : '&

# + - 5 *&

" : @ 5
5 8& 55< 4' 5
9 ; 5.4 0 .@ '? "1 ,)2

;6 =& =4 & B & * >) ! 2

< .4 2 4 9 4 ; ' 5= H + 6
H " 4 D ( A # 4 5 &

; 9& * '
.4 H + 6 B 5
5 < <&

; 9& = : =& * ' 7 < /

' ) @ ' 5D 5
9 4 . 5> 9 &

; 9& 8 & ;& = : =& >

. 4 5<6 = + < = - --5 H# B ; 6A , 6 6
5&

8 4 & ,,2 *' + +5

& 2 & &

171
References

8 @& & ,, ' ! . 5 <> ,,

8 @& - & ,, > ) >) - ,,

8 &D& < B & & ,22 >) <

6 - 9 < 6 <' < = 59 <
' 5= 5 8 & --, ,A

8 <4 &"& B4 45
5 &8& ,A 7 ! ,4 ' & . 4& = <&
;&'& &'&.& & -& !&

8G 4 < .& :?& " ! 9 -

0 .@ '? *,

@ 5 & D < @& & 5

5 D& "
< 4 )) ' 56 - %
! &

@ "&8& ,2( ; 5 . 5 < 6 9 <" <

9 4&

@ -- D&@& - 9&0& ,,A > />

" ' # *A (!,* &

D4 8&B & B 5
: @&;& ,,2 @
A ? , ;51 5 ' 5
5<
9 <= 5 H + 6MD 5
6 U) 5 % < H &

D4 & % 4 8& @ .&

6 @ > & >
" - ; <= > "1- % & )&

? : 4 D& J4 < & J '& 3 ! ' " 3 ,

! + ) D 5 ' 5
5< 9 < ?5
5 4 -)2 ( ,*A ** *

? : D& ,, 6 2 .4 ' # , ! ) !*!

=4 5 H + 6 . 4 5<6

? < & 6 9& . & ((

N' 5= - 5A 9 6 #
%5 &

172
 References

? 45 @& , ! '
' !0 .4 # & ,2 0 H ."? 0'= 7*) -
9 4 5" < < 0 :G- < H + " *2 2) 0 :G- < ' #,
!)!) (A 2

? 6: & ? 5:6 D >) ! ' !' )

' '' N " - ; <= > "1-

?65 & 0 & B & ,,! ' ' 3 " ; 4

' 5B : 4 - 5 5< % ) !

0 < ;& @ .& FG :5 & N " -

; <= > "1- % & )&

0 6 & ,,! 5 47 5 < < A4 9 5 '-

@ 5 5# D 6&

0 < & ' 0 .@ '? "C 22 &

0 V& 4 & =4 C& ,,A .4 < ;51 5 ;1 >B : 5 5

5
6E = /= 9 ; 4 ; < )&

0 S+ & " : '& B 5 9& * 3 '

'' < 4 - = < =" - (&

9 4 & ) ' * 7
= - > =4 < 5
&! # ) D - -- &

9 4 & 4 & 4 7
/ ' 7 0 ' = H .
= % 5 - # & A&

9 & & < D& 9 I D& 5

5 D& D I & 9 5 ;& 8 P I ;&
B ! $> " -
; <= > "1- % )

9 : & ,,2 * * * & "

/ 0 8 6 & &

9 & =W P 8& 9 < 6& 9& =& ( ' ,

' ''
" - ; <= > "1- - )&

9 : =& ' " - ; <

= > "1- % )&

173
References

9 6 0& * = + + 5-4 49 6 79 6 0& : ; ::

-5 *

# < & 56 8& % & B5

5 & = D& 0
5 , 5 " -
; <= > "1- &

# 4 .& V ?& ? V& @ & ,,) ! '

(4' & 6 -& ' 5 . :6 &

# ?&"& ,22 X NH& & (!) * * 9 & &

# 6 D& * = + < + 7# 6
D& @ : @ +5
5 D 69 5 5&

# 4 5B & 7 0) " -
; <= > "1-

# : 5 8& ,, 9 ' ' /@ ) <

" ,& 6 + . 4 S --5 & - < (* (*2
D = 8 &

# D& J4 < 9& ' " 3 ' ! =5 4<4

4 5<6 9 5# + &

%5 @& & ,, >) ? " '

< 4 ,, '""" ' 5=
9 6 ) 2&

% D& " 4 + 5 <7 + < 5 : - 4 5<6

4 5 + 4 9 H + 6 ; 56 = - <
" < <0 H ? <

"& " : '& J : 0& ,,! 0

' = - 2 - ? <5
. : : @G< : 5 :4 5

5
5 : @& ,!A ' # 2!, , 2( A

& ( *' 4 ( > 3 "

9 .4 0 .@ '? (/ O "&

4 & 9 4 / '
7<< ! 4 - 8 < -46 H + 6
5 5 #V ( A H &
174
 References

5 ?& = & ,,, ! '

<
2 ; <= > "1- % # 4+5
5
.# H &

5
5 & # +< & ) 1 '' 1 0
' #7 ) 2 )

(B " C ) !" ! + : = * ) )* + <

( ) )&

4 & ,, D *' B 4 ' > 0 " B >)

> 4 .4 . 4 4 H +
@ <@ <&

4 & 7 0 B > " -

; <= > "1- % )&

5 9&#& 8 6 %& "& 4 & ,,! >)

,? >) * + ' & D& + 9 < . 4 5<6 5
&
) # & , --& A A !&

.& 6 8& ) >) 0! '

9 ; 5.4 0 .@ '? "C 2

& ,A* 3 ) > &' 9 4 " <0 5

&
2 # * --& )! )2A&

& & ( >) " - ; <=

> "1- - )&

&
) 0 : 5 5
69
; 5.4 0 .@ ' V "C )(! ) 0 :G- < +

. -- 6 D&=& 0 =& & ,, & ! >) .4

' 5D 5 + 9 <. 4 5<6N&

G 5 < .& 0 S+ & * :3 ' > " ,

9 ; 5.4 0 .@ '? "1 */ "

B & =4 & ; 6 =& 8 6 #& 9 ? =& ) ;51 5

+ < ; 5< : 5E ; <= > "1-
- 9 5L= = # &

B4 "& #6 9& 8
" - ; <= > "1- &
175
References

B 5 (? % 1 2 " 2> 2
! ' ## &, 2( * H # &

B <4 ?& 4 & ,2, = - < L 56 B : 5

5
- 9 < - 9 < 9" ; 6&

B 5& & ( B &3B '' 3 >

' 0 .@ '? "1 )A&

V . ?& 4 D&D& 5: &D& ,22 '

) < 6 - 9 < 6 < ' <
= 59 <= 5 8 & --& 2

V . ?& < D&@& ,22 ,' ' )

< 6 - 9 < 6 < ' < = 5
9 <' 5= 5 8 & --& ) ,

J + & & < .& <5 & 4 & ( 3 3 ! ?

D, ) " ( - 9 <> ; <= >
"14 - 0 9% H &

176
 Appendices

Appendices

Contents Page
A: Adjusting a Hexapod to High Accuracy 178

B: The Hydraulic Diagram 179

C: A Short Summary of Transform Equations 180

D: The Force Cone Concept 182

177
Appendix A: Adjusting a Hexapod to High Accuracy

Appendix A: Adjusting a Hexapod to High Accuracy

Iteration
stops

Figure A1 : Adjusting a Hexapod position with the robot to 0.05 mm

Iteration
stops

Figure A2 : Adjusting a Hexapod orientation with the robot to 0.03

178
 Appendix B: The Hydraulic Diagram

Appendix B: The Hydraulic Diagram

179
Appendix C: A Short Summary of Transform Equations

Appendix C: A Short Summary of Transform Equations

The theory to this summary of transform equation is taken from Craig (1989).

The information needed to completely specify a location in the Cartesian space is a position and
an orientation. The situation of a position and an orientation pair arises so often in robotics that
we define an entity called frame, which is a set of four vectors giving position and orientation
information. A position P is described in a 3 x 1 position vector in coordinate system {A}
according to:

px
A
P = py
pz

The orientation {B} is described in a 3 x 3 matrix relative to coordinate system {A} according to:

r11 r12 r13

A
B R= [ X
A
B
A
YB A
]
Z B = r21 r22 r23
r31 r32 r33

Moving from one coordinate system to another is performed by multiplying one frame with a 4 x
4 matrix called a homogeneous transform. For example if a point P in coordinate system {B}
need to be described in coordinate system {A} the homogeneous transform enable us to go from
{A} via {B} according to:

A
P = BAR BP + APBORG

B
{A} Z B P
A
Z A P X B {B}

A
PBORG
YA YB
X A
Figure C.1: General transform of a vector

A
In this context PBORG is only the translational distance between A and B that is added to the
rotation transform of B P . In this dissertation the homogenous transform, moving from {A} to
{B} is further described according to ABT . Transformation of frames and coordinate systems can
be derived by the transform equation.

180
 Appendix C: A Short Summary of Transform Equations

{D}
{A}

{U}

{B}
{C}

Figure C.2: Set of transforms forming a loop

Figure C.2 is showing a set of transforms. There are two ways to go from {U} to {D}. First,
DT = AT DT ,
U U A

but also as
DT = BT CT DT .
U U B C

We may set these two descriptions of UDT equal to the transform equation:

T DAT =UBT CBT CDT .

U
A

Taking the inverse of a transform matrix is the same as changing the direction of the transform.
We have:
A 1 B
BT = AT .

This is used if the transform equation to solve for example is CBT . Using the inverse the solution to
B
CT is:
B
CT =UBT 1 UAT DAT CDT 1.

181
Appendix D: The Force Cone Concept

Appendix D: The Force Cone Concept

Load cells ABB IRB 4400 Robot

Cone

Figure D.1: The Force Cone developed by the author and University of Wollongong,
Australia

182
Appended Publications

Appended Publications I-VIII

183
 Appended Publication I

Kihlman, H., (2001) Reconfigurable Tooling for Airframe Assembly A state-of-the-art

Review of the Related Literature and a Short Presentation of a new Tooling Concept,
CIRP 1st International Conference on Reconfigurable Tooling, May 21-22
 RECONFIGURABLE TOOLING FOR
AIRFRAME ASSEMBLY

A State-of-the-art Review of the Related Literature

and a Short Presentation of a New Tooling Concept

Henrik Kihlman
Department of Mechanical Engineering,
Linkping University, Sweden
Telephone: 46-13-288974
E-mail: Henrik.Kihlman@ikp.liu.se

ABSTRACT

From the early days of aircraft manufacturing Dedicated Tooling has been used in the
assembly process to ensure the attainment of assembly tolerances and product quality.
Dedicated Tooling clamps the aircraft parts to be assembled into the jig to enable assembly
by riveting. However, increased competition in the aircraft industry has driven the need to
improve quality while reducing cost and in turn the need for innovative solutions to
accomplish this.

In this review paper the possibility of using metrology to increase the position accuracy in
robotics will be examined. This is necessary to be able to use robotics in assembly of aircraft
parts with the appropriate accuracy. Also, because of the small product volumes in the
aircraft industry, the jigs must be flexible in order to assemble more than one structure in
each jig. Solving these two problems could be the break through for starting to use robotics
in aircraft assembly at a higher rate, and doing so in a cost-effective way.

By then reviewing literature of todays flexible tooling technology in the aircraft industry,
the conclusion indicates that there is a gap to fill in aircraft assembly tooling. Modular Tools
is one solution where standard aluminium profiles are used to manufacture jigs with some
degree of flexibility. Another way is pogo fixturing, which uses sticks to hold airframe parts
together in the assembly process. The sticks can only be reconfigured in a limited range, and
are not cost-effective. By using Affordable Reconfigurable Tooling, the jigs will not only
have greater ability to be reconfigured, but by using robotics for the reconfiguration task as
well as for drilling, riveting and other material handling tasks, the system will also be cost
effective.

Keywords: Reconfigurable, Airframe, Assembly, Tooling, Flexibility, Affordable

1. INTRODUCTION

Today the aircraft industry uses jigs for aircraft assembly, designed through a traditional
technique called Dedicated Tooling. The jig supports components and acts as a positioning
gauge. Every assembly in the aircraft has its own jig, which is designed to be suitable only to
assemble that particular structure type. When the total batch of structures is completed, the
jig is either stored for future use or discarded. No material or functions from the jig is reused
or recycled. With only some exceptions, building aircrafts this way has been done for 60
years, and is safe and well known.

However, Dedicated Tooling is expensive and has long manufacturing lead times. It is also
not designed for variations in temperature, deflections or unstable foundations. Any tooling
realignment is time-consuming, and can often be achieved by the fettling and shimming.
Also the total product volumes in the aircraft industry are often smaller, seldom over 1000
(Boeing, 2001). In order to reduce cost and to increase affordability despite the small batch
sizes, there is a growing requirement for tooling to be flexible, reconfigurable and
reconstructable - flexible so that it can be quickly adapted to small variations within a single
structure, reconfigurable so that it can be changed within a structure family and
reconstructable so that it can be changed between different structure families.

In the aircraft industry today, technologies exist that have flexibility in the assembly process.
One is Modular Tooling, who uses standardised profiles to build the jigs, which makes it
possible to re-cycle the parts, and as the parts in the jig are not welded they can be adjusted
and thereby provide some flexibility. There also exist techniques to achieve reconfigurability
by use of so-called pogo sticks, which can change the configuration to reconfigure between
different airframe structures.

Flexible aluminium fixtures however, are more suited for adopting minor changes in the
airframe structure and must be re-built to handle other airframe structures. The pogo
fixturing technique on the other hand has the ability to reconfigure itself, but only in very
restricted manners. The Affordable Reconfigurable Tooling shall be able not only to handle
larger reconfigurations, but also in a cost effective way.

There has been very little research addressing the level of flexibility, which is about
comparing different concepts with different amount of flexibility. In this study a framework
for analysing different concepts of jigs with different level of flexibility will be presented.
This report of state-of-the-art will start out with an introduction to flexible tooling and move
on to the enabling technologies that is needed to make Affordable Reconfigurable Tooling
work.

2. AIRCRAFT ASSEMBLY TODAY

Flexibility in the production area and flexible production systems is being spoken of since
the beginning of the 80th. One clear reason for this is that most manufacturing industries
more or less have experienced the rising number of product variants with simultaneously
decreased product lifespan. Nowadays, more things shall be done in smaller quantities,
which increase the demands on the production systems. The system shall handle more
product variants, with differing customer demands and the ability to change product
generation frequently. The consequence of all this is that if a company uses product- or
variant specific tools, the cost needs to be distributed among fewer products, which gives
heavily increasing product costs as a result. George N. Bullen from Northrop Grumman
Corp ones gave an example of that big cost drivers are involved in aircraft assembly. He
reported that assembly-related operation account for over 40% of total airframe
manufacturing costs (Bullen 1999).
Traditionally, the design of an aircraft is also used to produce the aircrafts assembly tools
with which to manufacture that specific aircraft. Initially, master tooling gauges are
produced to an extremely high standard and calibrated using national standards. These
gauges are used to calibrate the fixed jigs and tooling which in turn are used in the
manufacturing process. The gauges form a physical standard to which the aircraft is
manufactured. However, there is a significant lead-time associated with the manufacture and
calibration of the tools. Jigs and tooling constitute a significant proportion of the cost of
manufacturing aircraft structures. The commercial aerospace sector is intensively
competitive manufacturers attempt to drive down the acquisition and operating costs of
their aircraft (Gooch, 1998).

Almost all the parts of an aircraft are assembled by riveting. However, currently the riveting
processes are done mostly by human operators, especially in small or middle scale
companies of aircraft manufacture. Most of the existing automatic riveting systems for
airplane fastening is very big and very expensive. In addition, the jigs for airplane assembly
are high in cost (Li et al, 1996). The cost of designing and fabricating the diversity of jigs to
satisfy the jig requirements of a manufacturing system can amount to 10-20% of the total
system (Nee et al, 1995), and the storage of dedicated jigs occupies lots of space. Jigs are
critical in the development of new manufacturing techniques and largely dictate the level of
flexibility a manufacturing system can achieve.

3. EXISTING FLEXIBLE AND RECONFIGURABLE SOLUTIONS

3.1 Modular Tooling

Modular Tooling is a tooling technology using modular thinking to get flexibility. The
modularity is about building fixtures from a collection of standard parts. The details can be
attached with ordinary screws and have some kind of slots, which enables the parts to be
adjusted and therefore becomes flexible, see figure 1. Aluminium is the most commonly
used material in these collections. Besides the flexibility, it is often possible to recycle these
standard parts, which enables parts to be reused for the next generation of fixtures.
 Figure 1: Extruded aluminium profiles connected with a common screw

As with all modular systems, the interfaces between different parts in the system are the
most crucial features. For a jig with the purpose to support, position and control an
assembly, the parts of the jig can be divided in three different categories the frame, the
adapters (positioning and holding the parts) and the links between the frame and the
adapters, who generally are called the pick-ups.

There is a lot of effort put in research projects, to develop effective modular assembly jigs.
The jig parts easiest to standardise is the frames, who need to have flexible interfaces
towards the pick-ups (this is one reason why extruded aluminium is interesting). To
standardize the rest of the jig system, there are two main tracks to follow:

A small number of standard adapters, which sets requirements on the details to have
standard interfaces
A small number of standardised pick-ups, who need to be a six degrees of freedom
system to overlap the distance and angular differences between the frames and the
adapters for all possible configurations

Advanced measuring technology together with computer support, simplifies building

methods like Modular Tooling. Using a metrology system for the calibration of fixtures
opens up for the use of digital master tools as well as for simplifying the build process itself.
Volvo uses CMMs and digital master tools to configure the jigs. However, it is also
possible to use measurement systems with more mobility, as photogrammetry or laser
tracker systems. Important to understand is that the general idea with Modular Tooling is not
about the use of extruded aluminium profiles, but for the idea of using modular thinking.
The reason why aluminium often is used in Modular Tooling is because the suppliers of
aluminium parts often offer the standardized collections of profiles and features to enable the
building of fixtures with standardized parts, and in a flexible and modular manner. Modular
Tooling has proven to reduce the development costs of the fixtures, as well as the time to
develop them.

One factor, which is often forgotten in the manufacturing industry, is that there are
environmental advantages to reuse production equipment. In fact, the European Community
is about to set new demands on the industry in Europe, to force them to have control over
what is brought in to the country and how to handle material waste, in order to more
effectively use our natural recourses.
 3.2 Reconfigurable Tooling

A way to achieve Reconfigurable Tooling is through the use of pogo sticks, see figure 2.
One of these applications is developed by Kostyrka Ltd. (Kostyrka, 2000). They use flexible
sleeves made from a compound of metal and plastic, which are axially held in housings. The
sleeves surround the part and clamp it by applying hydraulic pressure to the sleeve jacket.
These pogo sticks can be moved from one position to another either actively or passively.
The active pogo sticks are individually adjustable and programmable by their own controller
and servos to conform to part shape. The passive pogo sticks consist of actuators only
capable of extending, retracting and clamping. These pogo sticks are positioned through
external means, such as a robotic gantry or the machining centre itself. The gantry system
then sets the pogo sticks to the correct heights. The pogo sticks are placed in a fixture bed,
where they are positioned in a matrix. There is a vacuum cup on top of every pogo stick and
by the extraction of each pogo stick, they together can form a pattern and can hold plates and
skins with varied configuration.

Figure 2: Pogo sticks

Another company using reconfigurable tooling techniques with pogo sticks is

TORRESTOOL from mTorres in Spain. Their tool is a universal holding fixture, conceived
to support aircraft components in space. It may be arranged as horizontal, vertical or round
configurations, either active (servo driven) or passive (pneumatic) type of motions. Mostly
the Reconfigurable Tooling is used to hold plates and skins for trimming, drilling and
milling

3.3 Jigless Aerospace Manufacturing (JAM)

Another approach reducing the cost and increasing the flexibility of tooling systems for
aircraft manufacture is Jigless Aerospace Manufacturing (JAM). This approach strives for
the minimisation of product specific jigs, fixtures and tooling. A new integrated
methodology has been developed, which uses a number of building blocks and tools, to
enable design for jigless assemblies as a result of a logical, step-by-step process (Naing,
2000). In the traditional way of building aircraft, previously described as Dedicated Tooling,
the parts are located on reference set jig location. These jigs are dedicated to one assembly;
therefore they have no influence on flexibility. The parts held in the jig, are drilled and
fastened manually and deburring is required. This is a very labour intensive process. By
using JAM instead, parts may be assembled as part-to-part, where two mating parts are
drilled in isolation from each other and deburred. The holes in the parts are then used to
locate one part to another. No jig locations are used. Jigs only function as support cradles,
therefore giving flexibility. No deburr operation is required and the process will be less
labour intensive. With this technique the final position of the parts in the assembly is defined
in the detail manufacturing process. This on the other hand is sets higher demands on the
manufacturing process, where the risk of mismatch from one part to another increases.

Another way of using JAM is done by virtual reference. Here, a robotic arm holds the part
together. No jig location feature is necessary, therefore it is flexible and if enough pressure is
applied when drilled then no deburring operation is necessary. This technique sets high
demands on both the labour and the assembly system (Engstrm, 1998).

The Boeing Company has made an approach similar to JAM, called Determinant Assembly
(DA) (Williams, 1998), (Swanstrom & Hawke, 2000). With DA they can reduce the number
of tool-located components, by using coordinated fastener holes. All coordinated fastener
holes are drilled under size during part manufacture. The parts are then determinately located
through the use of coordinated fastener holes.

4. METROLOGY SYSTEMS

To be able to move from the old tradition of using Dedicated Tooling to the flexible tooling
technique, both the tools as well as measuring system need to be changed. When
determining accuracy of robotic arms, accuracy is separated into two categories, repeatability
and positional. Today industrial robots have fairly good repeatability accuracy, about 0.1mm
sometimes even better (ABB, 2001). The positioning is much worse. Most robotics
manufacturers do not even mention positional accuracy when performance is specified. This
is not an attempt to cover up some weakness; it is simply not considered a very interesting
factor in most robot applications. This partly due to the way robotic arms are traditionally
programmed, namely by teach-in. The teach-in method has the advantage that the
positioning error is compensated for. If the arm is repeatable and the work-piece is placed in
the same location in front of the robot, the end-effector will be able to perform its task in the
right place every time, despite the fact that the location for this task is more or less
unknown (Whinnem, 2000). When the process uses offline programming and not teach-in
operations, the positioning accuracy of robots is not enough. In the aircraft industry in
general the fixturing devices that hold parts together when building aircrafts must have
position accuracy better than 0.2 mm. A drilled hole in an airframe must have even better
accuracy. Some external measuring device is necessary to get the accuracy needed to drill,
rivet or assemble any aircraft structure. Today different measuring technologies have been
brought about to handle this problem. One of them is Photogrammetry and another is Laser
technology.

4.1 Photogrammetry

Richard Gooch made an interesting description of photogrammetry. He advocated that the

photogrammetry uses the known position of several camera stations together with the
projected angles of rays passing from the images of targets detected on the image plane of
each camera, to establish the 3D location of targets by determining the point in space at
which these rays intersect (Gooch, 1998). Richard explained the difference between Optical
Metrology and Machine Vision. The Optical Metrology uses high contrast optical targets
that provide the highest possible measurement accuracy, thus the need for sophisticated
image processing is eliminated in optical metrology. Machine Vision needs high performed
computing to extract and recognise natural features of an object. The goal of Optical
Metrology is to reduce costs and lead times while improving quality. Other important factors
are the elimination of Dedicated Tooling and increased level of automation. Richard
advocates that optical instruments containing advanced electro-optics, embedded processors
and digital interfaces are opening up new horizons for the application of optical
measurement in manufacturing automation. Metrology is an enabling technology and
applications are being extended from inspection to control of manufacturing process itself.

Another system that uses photogrammetry is 3D Image Metrology that has made large
improvements over the last years, particularly in three areas. The first area is quality control,
where production personnel and/or specialists operate the systems. In the second area,
machine control consisting of black boxes, which provide 3D positional feedback to CNC
machines or robots. The third area is in-process inspection, where the systems are integrated
to CNC machines, robots or production lines and perform measurements on the fly (Beyer,
1999). One system that uses the 3D Image Metrology is the TI2 technology at The Boeing
Company. This approach is based on the idea of controlling the location of a drill bit directly
in relation to the part using 3D Image Metrology instead of relying on expensive mechanical
systems. The TI2 system consists of the Tricept robot from Neos Robotics, the Imetric 3D
Image Metrology system from Imetric and the IGRIP simulation software from Deneb. The
TI2 system uses a 3D Image Metrology to inspect hole locations, trimming paths, and other
machining operations on the fly. But, the measured data is not directly fed back to update the
machine path in this system but to provide information for an inspection report and
statistical process control (Beyer, 1999)

4.2 Laser Measuring

The rear structure of the Airbus A340-600 has been assembled with two laser
interferometers, manufactured by Leica. According to the expertise the traditional
construction tools could be replaced by supports to hold the work pieces together in space.
The laser trackers use the angles derived from the virtual CAD model to measure the real
object. Consequently the work pieces can be positioned in real time to each other using
reference targets on each part. Thus all sections could be adjusted immediately; the tooling
cost could be cut by half. The laser tracker positions all work pieces to be assembled. Earlier,
the measurements had to be taken manually, now they are generated automatically (Leica,
1999).

As mentioned in section 3.3, The Boeing Company has done an approach to achieve flexible
tooling, by using Determinant Assembly (DA). This technology results in a flexible and
more accurate assembly system. DA eliminates the need for master tooling gauges by
building the jigs and tooling to CAD and calibrating these directly using optical
measurement systems such as the laser tracker. The changes to assemblies can be
accomplished by modifying the feature locations in a part NC program. This is contrasted to
Dedicated Tooling where a physical component of an assembly fixture must be relocated or
a new index fabricated and installed on the assembly fixture (Williams, 1998).

5. A NEW TOOLING CONCEPT

The relationship between the existing tooling concepts today and the Affordable
Reconfigurable Tooling concept is illustrated in figure 3. It is positioned in between two
technologies. To the left are tools for aircraft assembly, and to the right fixtures working as
holding devices for aircraft part manufacturing. This section will briefly describe the five
different tooling concepts. They will be compared through their ability to reconfigure
between different configurations in a short period of time.

Figure 3: Five Tooling Concepts with different ability to reconfigure and different degrees of
flexibility.

5.1 Assembly Jigs

Concept nr. 1 in figure 3 is called Dedicated Tooling. This concept is the most commonly
used tooling technique in aircraft assembly today. Because Dedicated Tools are tailor made,
they have the ability to assemble all kinds of airframe structures. Every tool is designed to
assemble one particular structure. On the other hand if changes are required, the jig has to be
sawed apart; new parts designed, manufactured and finally welded or perhaps screwed into
the jig to the right position. The consequence of this is that this kind of assembly jigs never
is reconfigured.

Concept nr. 2 in figure 3 moves us to Modular Tooling, which is built on standard

aluminium profiles, see section 3.1. Those jigs might be changed, but in a limited range. By
using pick-ups to adjust to the datums (e.g. fixturing points), those jigs may handle minor
changes from the original configuration. They are flexible but not very reconfigurable.
Reconfiguration between different airframe structures is hardly an option. But suppose that
the Modular Tools has to be reconfigured, to assemble another structure type, the jig would
first have to be disassembled and then rebuilt again to apply the new configuration.
Compared to Dedicated Tooling, the Modular Tooling has advantages. Most of the jig
structure is re-usable, the parts are standardised and the jig is adjustable, but not very
reconfigurable.

5.2 Manufacturing Fixtures

Concept nr 3 in figure 3 is the Pogo-fixturing concept, which is previously described in

section 3.2 as reconfigurable tooling. This solution is today mainly used to fix airframe parts
for milling and drilling. The fixture is most often used to suck plates with vacuum cups, so
that a gantry robot can mill the edges on the plate or drill holes. Although these fixtures are
most commonly used for fixturing details for manufacturing there do exist applications
where assembly is done. One example is the assembly of stiffeners to plates. Although this is
assembly, it has a very restricted change over rate. The ability to reconfigure between
different airframe structures is possible, but rather small.

Concept nr 4 in figure 3 shows a concept called Hyper-flexible concept. This solution is used
mostly for holding parts for manufacturing (e.g. flexible NC-fixtures). The change over rate
is rather small, but the time to reconfigure is fast. This technique is old and well known and
very much developed. Although the NC fixtures may reconfigure quickly, they have
geometrical limitations in the ability to change over between different types of structures.
They are most commonly used to fix smaller details for manufacturing. No assembly is
involved.

5.3 Affordable Reconfigurable Tooling

By using the ideas from reconfigurable manufacturing fixtures combined with the ideas from
the modular thinking, a new concept has been developed. This concept is called Affordable
Reconfigurable Tooling. Basically this solution uses an industrial robot to do the
reconfiguration task. The jig will need some kind of pogo sticks, similar to the passive pogo
sticks from Kostyrka (section 3.2), although they will almost certainly need to be modified.
The pogo sticks have some kind of locking device, which is inactive in order to be
reconfigured and active to be fixed. The reason why they are locked when the system is
deactivated is because there is always a risk of leakage with a pressurised system if the tool
is not reconfigured in longer periods of time.

Although the assembly system goes from being dedicated to reconfigurable, there will be
limitations in the change over rate that the system will manage. Perhaps the system has the
ability to reconfigure within one family of structure types (e.g. planar structures, wing
structures or aircraft bodies). This might be enough for some aircraft assemblers who have a
specific niche on the market, for example medium sized wing structures. But if there still are
demands to handle a reconfiguration between product families the need to reconstruct the
tool, the pogo sticks should be modular in order to making it possible to reconstruct the tool
to a bigger change over than the pogo sticks will manage. Because the pogo sticks are
modular they can be dismounted and applied in some other configuration. This will probably
be done manually and therefore take more time, but perhaps that is acceptable for some
assemblers, where longer reconfiguration time is acceptable.
To clarify the distinction between Pogo-fixturing and Affordable Reconfigurable Tooling
one can say that in the latter case there is the advantage to use a robot in the reconfiguration
process. It probably will be cheaper to buy passive pogo sticks. Their only task is to be
flexible when unlocked and rigidly fixed when locked. No expensive equipment is needed in
each pogo stick to give the high accuracy in positioning. By using a robot to do the
reconfiguration task, no built-in servo is needed in every pogo stick to move it. Investing the
money in a robot will probably be the cheapest investment in this comparison, not only
because of the cheaper pogos required, but also because of possibility of using the robot to
do drilling, riveting as well as other material handling tasks.

6. CONCLUDING REMARKS

In section 5, a short introduction was given on how Affordable Reconfigurable Tooling may
be a way to fill the gap between todays assembly jigs and part manufacturing fixtures. The
question is probably how much time we can afford for the reconfiguring process. The
University of Linkping in cooperation with nine other Airplane manufacturers and
suppliers, are now under way to continue this work, which also is founded by the European
Community Automation for Drilling, Fastening, Assembly Systems Integration, and
Tooling (ADFAST). Still there is a lot more work required in the research of how the
Affordable Reconfigurable Tooling system shall be designed to accompany a wider range of
reconfiguration, robustness and accuracy.

An other interesting area, where a lot of progress has been made recently is in the metrology
area. Optical measurement sensors are increasingly available, often finding application in
measurement and inspection of manufactured products. For example, theodolites and laser
trackers are already used to calibrate jigs and tooling. Digital photogrammetry is used in
dimensional inspection of assemblies such as aircraft wings. Such tasks demand high
performance sensors with 2D and 3D capability, large working envelopes, high accuracy,
low measurement latency and increased flexibility. The availability of sensors, which meet
and exceed such criteria, is fuelling new possibilities in the manufacturing process itself.
Dedicated Tooling may be eliminated and replaced by Affordable Reconfigurable Tooling
under the control of embedded sensor systems. But a lot more research is required before the
accuracy in the machine controller is to perform with enough precision and speed.

As the interaction of 3D models and 3D metrology is making it possible to close the link
between designs and manufacturing, the vision to start using virtual manufacturing, which is
about going from CAD solids to accurate assembled aircraft structures is coming to be a
reachable strive. But considerable effort is needed in the reconfiguration programming
process in order to shorten the long lead times in Dedicated Tooling. The time and effort
must not be translated in complex offline programming procedures and end up in continuous
long time and thereby high costs. The offline programming system needs further
development in order to function as an operation planning system as well. This makes the
programming and operation planning of the system. Today there exists software for this kind
of processes, for ex. RobCad from Technomatix, or IGRIP from Deneb. If the kinematics of
the tool is defined in the offline system, simple drag and drop methods is one easy way of
solving it. But still, a lot of work and effort is needed before we get there.

A lot of work is still left to be done to achieve Affordable Reconfigurable Tools, but if they
become a reality, industrial robots will finally find their way into the aircraft industry, and
reduce the labour intensive assembly process as well as drilling and fastening, which would
make the aircraft manufacturers more Lean, Agile and Flexible.

8. REFERENCES

ABB, 2001. Product brochure for IRB4400 3HAC 10527-1 7 M2000 from ABB Flexible
Automation.
Beyer, H.A., 1999. 3D Image Metrology for Lean Manufacturing. SAE Aerospace
Automated Fastening Conference & Exposition.
Boeing, 2001. Boeing commercial airplanes, 2001, Order Summary By Year -- As of
December 2000, http://www.boeing.com/commercial/orders/ordsumbyyear.html,
accessed 12/01/01.
Bullen, G.N, 1999. Assembly Automation and Implementation Issues. SAE Aerospace
Manufacturing Technology Conference & Exposition
Engstrm, M., 1998. Flexible Workshop For Airframe Assembly. Nouvelle Reveu
DAronautique et Dastronautique, No2 1998 3rd Aero days post-conference
Proceeding.
Kostyrka, P.A., Kowalsky, J. 2000. Flexible Active and Passive Pogo Fixturing Systems for
Aircraft and Aerospace Applications. SAE Aerospace Automated Fastening Conference
& Exposition.
Li, Y., Bahr, B., Chen, X., 1996. The design of a Flexible Fixture & Workcell for Aircraft
Assembly.
Naing, S., Burley, G., Odi, R., Williamsson, A., Corbett, J., 2000. Design for Tooling to
Enable Jigless Assembly - An Integrated Methodology for Jigless Assembly. SAE
Aerospace Automated Fastening Conference & Exposition.
Nee A.Y.C. Nee, K. Wyhybrew and A. Senthil Kumar 1995. Advanced Fixture Design for
FMS. Springer-Verlag Lindon limited.
Gooch, R., 1998. Optical metrology in manufacturing automation. Sensor Review 1998 vol.
18 nr. 2.
Leica, 1999. Journal dInformation Interne dArospatiale Matra Airbus, No 6, October 1999
http://www.leica-geosystems.com/ims/application/aerospatial_nantes_fr.pdf, accessed
25/12/00.
Swanstrom, F.M, Hawke, T., 2000. Design for Manufacturing and Assembly: A Case Study
in Cost Reduction for Composite Wing Tip Structures. SAMPE Journal, Vol. 36, No 3,
May/June 2000.
Whinnem, E., 2000. Integrated Metrology & Robotics Systems for Agile Automation. SAE
Aerospace Automated Fastening Conference & Exposition.
Williams, G, 1998. Gaugless Tooling. SAE Spring Fuels & Lubricants Meeting &
Exposition.
 Appended Publication II

Herbertsson, J., Kihlman, H. and Engstrm, M., (2002) "Reconfigurable Aircraft

Assembly Using industrial robots and new tooling to meet future production scenarios
", 33rd International Symposium on Robotics in Stockholm, October 8-10
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
Reconfigurable Aircraft Assembly Using industrial robots and new
tooling to meet future production scenarios
Jonas Herbertsson, Henrik Kihlman,
jonhe@ikp.liu.se henki@ikp.liu.se
Linkping university
Dept. of mechanical engineering
SE-581 83 LINKPING
SWEDEN
ABSTRACT
This paper presents the development of a new
concept for aircraft airframe assembly tooling. While
conventional aircraft assembly tooling relies on rigid
steel frames this novel approach can be re-configured
for different products by the help of an accurate laser
tracker guided industrial robot. Through this, aircraft
manufacturers can better cope with an increased
number of variants and smaller volumes which will
increase the already high tooling costs. So far in the
project a tooling concept has been conceptually
designed and economically evaluated. The early
technical results of the conceptual work indicate that
the concept will work provided that enough
mechanical stiffness can be obtained. The economical
analysis shows that despite a much higher investment
cost, a reconfigurable tool can be economical if it will
replace 4-5 conventional tools. In certain future
production scenarios reconfigurable tooling can play
a very important role to keep the tooling costs down.
Keywords: robot reconfigurable aircraft assembly
tooling.
1 INTRODUCTION
1.1 Background and problem
As in many other industries the aircraft industry also
faces new preconditions for manufacturing like e.g.
an incresasing number of variants, shorter
development lead times etc. Today, aircraft airframe
assembly is usually performed by the use of rigid
steel tooling (fixtures) dedicated for specific
products/parts. These require redesign and
configuration for each new product or assembly
operation, which is costly and result in long lead time
(4). According to Bullen, (1) quick set-up and flexible
tooling for major aircraft assembly is opening the
door to automation and mechanisation.
In aircraft assembly, a very high number of assembly
tools are used and therefore the tooling cost is both a
significant development and manufacturing cost.
The objective of this paper is to present a
reconfigurable assembly tooling concept that can be
used for different assemblies, thereby making it more
Magnus Engstrm,
magnus.engstrm@saab.se
SAAB Aerospace
Commercial programs
SE-581 88 LINKPING
SWEDEN
economical to manufacture airframes in smaller
volumes, more variants and other applications which
otherwise would have demanded new tools.
2 FLEXIBLE ASSEMBLY TOOLING
IN AIRCRAFT INDUSTRY
2.1 Earlier attempts
Attempts to develop more flexible aircraft assembly
tooling have been made sporadically since the 1980s.
Among these there seem to be very few that have
been based on the use of an industrial robot.
According to Wright and Sahr (3) work stoppage
resulting from process anomalies are significant cost
drivers in todays non-automated environment. Asada
and Bernard, 1984, claimed that cost and lead time
could be reduced by using a robot manipulator to
automatically change the layout of a reconfigurable
fixture system. Wright and Sahr, (3) presented an
example of a flexible assembly subsystem, which aim
was to develop and demonstrate the technologies
required for flexibility and adaptability in automated
assembly for this high-precision, low-batch-count
manufacturing environment.
Goodman (13), 1990 presented a flexible jig where a
gantry robot was used for reconfiguring the fixture
between different airplane products within a product
family. Goodman estimated that one flexible tool can
replace up to 250 conventional tools!
However, for different reasons it seems like flexible
airframe assembly tooling so far have not succeeded
to compete with conventional tooling. There could be
several reasons for this, for instance:
- The required tolerances cannot be obtained
without expensive NC-controlled equipment.
- The robots to be used do not have enough
positional accuracy for this purpose.
- Many airframes need so complex assembly tools
that a reconfigurable tool hardly can hardly be
designed due to the spatial limitations.
Today all three hinders are more or less eliminated.
By using a state of the art laser tracking system an
industrial robot can obtain enough position accuracy.
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
Also the aircraft assemblies are becoming less
complex in terms of number of parts and geometric
variation thanks to modern high-speed machining.
There will in other words be less but larger parts to
assemble which will make it easier to obtain enough
free space in the tools for robots and other equipment.
In other words it is now better opportunities than
before to succeed with reconfigurable aircraft
assembly tooling.
2.2 Possible tooling concepts for airframe
assembly
Flexible tooling in general has been used for a long
time, especially for fixturing of workpieces during
machining. A relatively new idea for assembly
tooling is the so called "modular tooling" where
extruded aluminium profiles are used. Such tools are
for instance used in car industry for sheet metal
assembly. Modular tooling gives you the abilty to
adjust the configuration of the tool by just loosen
some screws (5). This will enable the tool designer to
adopt for product changes late in the product
development. Tool parts of modular tools can also be
recycled for the next generation of tools to cut costs.
Another flexible tool solution is "hyper-flexible
tooling", (CNC-controlled tooling). This solution is
based on the ability to re-configure the tool in just a
few seconds so that different assemblies within for
instance a product family can be assembled or
machined. Normally this type of tooling is only used
for machining operations and not for assembly.
In 1990, Goodman (13) mentioned that future aircraft
assembly tools probably would be reconfigured by
the use of servos and encoders to perform the set-up
change of the fixture. Similar ideas were also
presented by Li et al. (7) in 1996.
Today there are several suppliers of this kind of
"hyper flexible" tool systems. One manufacturer of
this type of tooling is MTorres in Spain (11). Another
type has recently been presented by Kostyrka and
Kowalsky (6). They have also taken part in
discussions of the use of passive dynamic modules,
meaning that the movement in the dynamic
(movable) module would have to come from an
external source, such as e.g. a robotic device. The
dynamic module only withheld the configuration
between the set-ups through a built in locking
mechanism. This principal idea has also been studied
in an EU-funded project called "Flexible Workshop
for Airframe Assembly" described by Engstrm (2).
3 A NEW RECONFIGURABLE
TOOLING CONCEPT USING AN
INDUSTIRAL ROBOT
By joining some ideas from modular tooling and
hyper-flexible tooling based on the idea of letting an
external device position passive dynamic modules, a
new reconfigurable tooling concept is being
developed. This tooling concept is called an
"affordable reconfigurable tooling" (ART) and the
very first ideas have been presented by Kihlman (5)
and by Kihlman and Engstrom (8).
Since tool reconfiguration times not are critical in
aircraft assembly it is fully possible to let an external
device reconfigure each dynamic module in turn.
The new concept is based on the use of a standard
industrial robot to sequentially reconfigure each
dynamic fixture module. The robot is cost-effective
since the idea is to also use it for assembly process
operations such as drilling, countersinking, riveting
and materials handling. The dynamic modules of the
tool in turn are attached to a static framework and
they have built-in locking mechanisms to hold the
configuration between the set-ups. A principal sketch
is shown in figure 1.
Figure 1: A principal sketch on the ART concept
showing the static framework and the dynamic
modules which attach to aircraft parts.
The static framework represents a universal tooling
system together with the dynamic modules. Each
dynamic module can not only be re-configured with 6
degrees of freedom but also be moved to another
position on the static framework. This latter process
is here called rebuilding. Rebuilding will only be
performed if the current configuration of the tool
cannot handle the geometry of the next product.
Assembly analysis and programming will be done on
an offline simulation and programming system
(OLP). Such systems are normally used for the
programming of the robot but the kinematics of the
dynamic modules can be defined and simulated as
well. Figure 2 shows a model of a tooling with its key
components,
3.1 The re-configuration process
A simplified description of a reconfiguration process
is as follows:
1. Using the OLP, position the workpiece(s) inside
the static framework without any dynamic
modules.
2. Study the position and direction of all pick-up
points. Re-build if necessary.
3. Locate the appropriate positions of the dynamic
modules on the static framework to get access to
each pick-up point.
4. Make a FEM analysis of the tool configuration.
5. Simulate and program the robot trajectories
between the previous and the current
configuration of the dynamic modules.
6. Simulate the drilling and the riveting operations
for each stage
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
Figure 2: A model of the full tooling system with its
three major parts; the tool (with airframe parts), the
robot and the laser tracker.
When all dynamic units are configured by the robot,
the set-up of the universal tooling system is complete.
Next step is to put in the pick-ups. The pick-up is the
tool closest to the product (aircraft parts). These units
can be made very small and could either be of
standard types or dedicated for specific assemblies.
The pick-ups have interfaces to the dynamic modules
based on the Coromant Capto toolholding system
from SANDVIK. At the end of each dynamic
module a male part of the capto system is used. A
Capto chuck is positioned at the robot TCP at all
times. The same system is also used on the drilling-
and riveting machines which will be handled by the
robot.
In general, the accuracy for the position of a pick-up
point is approx. +/- 0.2 mm. Since the robot is
offline-programmed it is necessary to rely on the
positional accurcy of the robot. However, since
industrial robots on the market today cannot manage
this accuracy, a six-degree-of-freedom measuring
system will be used to improve the accuracy of the
robot. The system is based on a Leica laser tracker.
In order to configure a dynamic unit the robot docks
with the Capto-interface of the unit and moves it to
a desired position and orientation in space. When this
position is obtained, the measuring system checks if
the position of the robot tcp is well within the
required accuracy. The syncronisation time between
the laser and the robot is short, well enough to control
the robot postion during movement thanks to new
software with embedded objects integrated with the
measuring system.
3.2 Current status in the project
The work on this new tooling concept is currently
performed in the EU-funded project ADFAST (9).
However, a foundation to this research was also an
economical evaluation made in the project "Flexible
Airframe Assembly" funded by the Swedish
Aeronautical Research Programme (NFFP), which
ended in January 2002.
The EU funded project ADFAST, has currently
passed the mid-term evaluation and several new
insights into automation of aircraft production using
industrial robots have been made.
One work-package of ADFAST includes robotic
drilling and riveting and interesting results have been
obtained regarding a new drilling technique called
orbital drilling (10, 12).
In another work-package of ADFAST the
reconfigurable tooling concept has been further
developed.
A third work-package concerns the development of a
6 DOF laser integrated robot control.
Currently, demonstrators of the reconfigurable
tooling are being detailed designed and
manufactured. These will be tested and evaluated
during 2003.
4 ECONOMICAL PRECONDITIONS
IN FUTURE PRODUCTION
SCENARIOS
The possibility of succeeding with the introduction of
a new tooling technology is naturally very much
dependant on the economical properties. Especially
when a new technology must compete with an old
well known one it is very important to early estimate
whether the new technology actually can compete at
all. Functionally the new tooling has to do exactly the
same as the conventional so the benefit must be
economical. For these reasons an early economical
comparison between conventional tooling and a
reconfigurable tooling has been made in order to see
whether it will be any idea or not to continue the
development for economical reasons.
4.1 Possible applications of reconfigurable
tooling
Since a reconfigurable tool will demand a
significantly larger investment than a conventional
tool, the economy of it will first of all be based on the
fact that the tooling and other equipment can be used
for several products. In other words, the more
products that can use the tool and consequently the
more conventional tools that can be eliminated the
more economical will it be. An economical
comparison between conventional and reconfigurable
tooling therefore very much depends on the actual
application and the production scenario over a long
period of time. Thus, it is not possible to come to a
general conclusion whether reconfigurable tooling
always is economical or not. Instead an economical
evaluation can indicate how and under what
circumstances reconfigurable tooling may be
profitable.
The first step in the evaluation was therefore to
identify applications for reconfigurable tooling. Three
types of applications that can be identified are:
Product family assembly
For manufacturers of commercial aircraft and for
suppliers it is often the case that the same airframe
parts of different models are manufactured in the
same location. Suppliers also tend to become
specialists of certain airframe parts such as e.g.
horizontal tail planes and elevators. In such cases
reconfigurable tooling could be a way of reducing the
tooling cost per product by investing in a single tool
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
or very few tools instead of one conventional tool per
product variant. The economy of a reconfigurable
tool here lies in that it can be used for similar parts
of different products.
Small scale manufacturing
Some aircraft are produced in very small volumes
compared to for instance short-distance commercial
aircraft. This could be the case when production of
special military aircraft, certain commercial aircraft
models, flying demonstrators or even prototypes are
concerned. In those cases the total manufacturing
volumes could be between 1 to 100 units. In such
cases reconfigurable tooling can be very economical
if it is possible to use it for different airframe parts
of the same product.
Manufacturing ramp-up
A third interesting application is when the peacetime
production rate of a military aircraft is very low and
then suddenly a very high production rate is wanted.
In this application a reconfigurable tooling can be
both economical and a significant enhancement for a
fast and dramatic increase in the production rate.
A reconfigurable tooling will also make it easier to
continuously update the product since design
modifications are easier to perform.
4.2 Risks and weak points
Even if reconfigurable tooling is likely to have
economical advantages there are also some weak
points and risks that have to be considered during the
continuous development. A very critical technical
issue, stiffness, has already been pointed out.
However, this economical evaluation has been made
under the assumption that the basic functionality of
the tool (too locate and maintain the position of
aircraft parts during assembly) is equally performed
in both conventional as well as reconfigurable
tooling.
4.3 Economical comparison with
conventional tooling
As indicated above an economical comparison with
existing conventional tooling must be based on the
estimation of how many conventional tools can be
avoided by using the new technology. This number
will depend very much on the actual production
scenario over a long time of years and therefore it has
been necessary to look at a couple of such scenarios
in order to perform an early economical analysis.
However, before presenting the results of such
analysis the basic economical properties must be
described.
Basic economic properties
The most significant costs of aircraft assemly tooling
is the design and manufacturing of the tools together
with material used. These costs can be divided into
the categories of product specific costs or common
costs (not specifically related to a certain product).
Conventional tooling usually have mainly product
specific costs while the reconfigurable tooling aims at
having mostly common costs. However, even the
most flexible tooling cannot avoid the fact that certain
product specific parts need to be used such as e.g.
contour boards.
In this economical comparison the costs for
conventional tooling have been estimated by
gathering economical numbers from the real
development of a tool at SAAB Aerospace.
The estimation of specific and common costs for
reconfigurable tooling has been made by configuring
a tool concept for the same assembly as above and
trying to estimate the costs for the parts and the
design and the manufacturing of it. Since the concept
is based on the use of an industrial robot and an
advanced measuring system it has also been
necessary to handle these costs. This was solved by
simply estimating how much of the investment costs
of these that could be allocated to the reconfigurable
tool. Our estimation was that less than 20 % of the
robot and measuring system costs could be allocated
to the tool. In other words 80 % of the robot cost
should be allocated to assembly process costs.
The result of the estimation of costs is shown in
figure 3.
As seen in the figure a reconfigurable tool for the
actual structure will need almost three times the
investment of a conventional tool. Of this investment
a small amount is product specific costs that will
occur for every new product.
Investment costs for re-config. and conv.
tooling
800
700
600
Money unit
500
Common costs
400
Product specific costs
300
200
100
0
Conv. Flex.
Tooling concept
Figure 3: The relative difference in investment costs
between conventional and reconfigurable tooling.
A conclusion that can be made based on this first
relative comparison of investment costs is that the
investment in a reconfigurable tool must eliminate at
least 3 conventional tools in order to get a break-
even. However, such an analysis should also include
the fact that investments in conventional tools may
come several years later when new products are
introduced. This demands that effects of the
difference in capital flow and interest rate must be
included into the comparison which will give a
positive effect on the investment in conventional
tooling. This is also considered in the following
analysis of a production scenario.
Comparison based on a mixed production scenario
One way of using the reconfigurable tooling is to find
an application where similar airframe structures from
different aircraft are assembled in the same tool. In
the economical evaluation described here it is
assumed that each airframe structure has approx. the
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
same assembly content and each tool has the
maximum capacity of 200 units per year. A tentative
production scenario in terms of products and volume
over 15 years is illustrated in figure 4.
Produced units per product and year
500
400
Volume
300
200
100
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Year
Figure 4: A tentative production scenario with a
mixture of products in different volumes over a 15
years period.
As seen in the figure there is a mixture of more
volume products and products that come in much
smaller volumes and even prototypes or similar that
are manufactured in just a few units. The total volume
per year varies between 200 and 400 units which
means that two tools are needed for the capacity. This
production scenario will demand different
investments schemes as seen in figure 5 below.
Investment costs over a 15-year period
800
Investment
600
costs
400
200
0
Conv.
1
11
13
Re-config. Year
Figure 5: Investment schemes for conventional and
reconfigurable tooling over a 15-year period. (There
is no unit for the costs since it is the relative
difference that is of interest.)
A decision to go for the reconfigurable concept in this
scenario will demand two large investments in year 1
and 3 while conventional tooling would need a serial
of 8 investments all over the period.
Even if it is obvious that 2 reconfigurable tools here
will replace 9 conventional it is not obvious that it
will be an economical investment before looking at
the net present values (NPV) of both alternatives.
Such an analysis actually shows that the difference is
much smaller than the difference in total investments.
The NPVs will actually be 1345 (money unit) for the
conventional alternative and 1190 for the
reconfigurable. (NPV calculated based on a 5 %
interest rate.) With a higher interest rate it might even
turn out that the conventional tools will be cheaper in
the long run in this particular case.
Low volume production scenario
Another production scenario that has been
investigated is low volume production. In this case it
has been assumed that the production of an aircraft at
a rate of 20 per year is needed. If it were possible to
produce five major airframe assemblies in the same
reconfigurable tool this would mean that five
conventional tools could be replaced already from the
beginning. In this case a reconfigurable tool will be
the most economical alternative by far, especially if it
can be used also for other products. There is a lot of
I free capacity (100 units/year) as long as only 20 units
H per year of these five assemblies are produced.
G The economical analysis becomes more complex if a
F production ramp up scenario is combined with the
E low volume scenario. If for instance the production
D rate after 10 years dramatically increases up to a peak
C level of 600 units per year this means that two more
B reconfigurable tools are needed. If conventional tools
A were used from the beginning then 2 copies of each
existing tool is needed. The NPV of these two cases
gives the result that the two alternatives are fairly
equal in terms of economy. However, this is true only
if the reconfigurable tools are used solely for one
product. In reality, it would be possible and very
likely that the tools will be used for other products.
This will definitely make the reconfigurable tooling
alternative the most economical.
4.4 Qualitative economical factors
Besides this quantitative economical analysis there
are also important properties of reconfigurable
tooling of qualitative nature. For instance aspects that
might have a significant positive economical effect
are:
- Better possibilities to level out variations in the
production rate due to the increased flexibility.
- Less need for floor space.
- Easier and cheaper to make changes in the
airframe design.
15
- Possibility to very rapidly prepare and re-
configure for a new assembly since a lot of the
work can be performed in a computer using
CAD-models and off-line programming.
- A significant drawback of reconfigurable tooling
may be the risks the technical complexity could
lead to.
5 DISCUSSION AND CONCLUSION
A novel concept for a reconfigurable aircraft
assembly tooling that is under development has been
presented. The concept is based on the use of an
industrial robot for the re-configuration of the tool for
another product. By using a robot with high position
accuracy that will also perform assembly operations
the investment cost for the tooling can be kept low. In
order to obtain enough position accuracy the robot
will also be guided by a 6-DOF laser tracker system.
The tooling is divided into a static framework and
dynamic (movable) units.
Reconfigurable tooling could be an essential part to
meet new demands in aircraft manufacturing. The
economical evaluation of the new tooling concept in
different scenarios points out that in most cases where
a reconfigurable tool can eliminate 4-5 conventional
tools it will be an economical solution. In some
 Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002

scenarios reconfigurable tooling will definitely be a
very strong economical alternative with some very
interesting properties.
However, there are still many technical aspects to
analyse e.g. stiffness and the overall system
functionality (tool, robot and tracking system) has to
be tested in reality. Therefore, prototypes of
reconfigurable tools are currently being manufactured
for evaluation. This and other further work will
continue over the coming years.
In the end, this work hopefully will lead to an
increasing automation level in aircraft industry with
more industrial robots. And this at the same time as
manufacturing flexibility is increased in order to meet
the future production scenarios of aircraft industry.
(10). Kihlman, H., Eriksson, I., Ennis,. M.,
Robotic Orbital Drilling of Aeronautical Structures
and Materials, SAE Aerospace Automated Fastening
Conference & Exposition, 2002
(11). http://www.mtorres.com/ingles/ia/index
.asp
(12). Lindqvist, R., Ericsson, I. and Wolf M.,
"Orbital drilling of sandwich constructions for space
applications", , SAE Aerospace Automated
Fastening Conference & Exposition, 2001
(13). Goodman, H., "Assembly cell speeds
riveting", Journal of tooling and production, Dec.,
1990

6 ACKNOWLEDGEMENTS
This work was partly funded by the National Swedish
Aerospace Research Programme (NFFP) and by the
European Union via the ADFAST-project
(Automation for Drilling, Fastening,
Assembly, Systems Integration, and
Tooling) (www.euadfast.com).

7 REFERENCES
(1). Bullen, N., The mechanization/automation of
major aircraft assembly tools, Production and
Inventory Management Journal; Falls Church Quarter
1997
(2). Engstrm, M., "Flexible Workshop for Airframe
Assembly", Nouvelle revue daeronautique et
dastronautique, nr 2, 1998
(3). Wright, K., and Sar, B., Computer integrated
Quality Assurance for Robotic Workcells in
Aerospace Manufacturing, Robots in Aerospace
Manufacturing, SME, February, 1989
(4). Asada, H. and By, A. B., "Implementing
Automatic Setup Change via Robots to achieve
Adaptable Assembly", American Control Conference,
San Diego, CA, 1984
(5). Kihlman, H., Reconfigurable Tooling for
Airframe Assembly, CIRP 1st International
Conference on Reconfigurable Manufacturing, May
21-22, 2001
(6). Kostyrka, P.A., Kowalsky, J. Flexible Active
and Passive Pogo Fixturing Systems for Aircraft and
Aerospace Applications, SAE Aerospace Automated
Fastening Conference & Exposition, 2000
(7). Li, Y., Bahr, B., and Chen X., The design of a
Flexible Fixture & Workcell for Aircraft Assembly,
CAD/CAM Robotics and Factories of the Future,
Middelsex University, London, England, August 11-
13, 1996.
(8). Kihlman, H., Engstrm, M., Affordable
Reconfigurable Tooling, SAE , SAE Aerospace
Automated Fastening Conference & Exposition, 2002
(9). ADFAST, 2002. ADFAST project web site,
http://www.euadfast.com,
 Appended Publication III

H. Kihlman and M. Engstrm, "Affordable Reconfigurable Tooling", SAE 2002

Transactions Journal of Aerospace, ISBN: 0-7680-1285-6

Affordable Reconfigurable Tooling
Copyright 2001 Society of Automotive Engineers, Inc.
ABSTRACT
Since the early days of aircraft assembly, welded steel
structures called Conventional Tooling has been used
for positioning and holding parts in place during
assembly. This paper presents a new tooling concept
called Affordable Reconfigurable Tooling, where a robot
is not only used for drilling and riveting but also for
reconfiguring the tool itself. The concept consists of
modular units that can either be reconfigured between
products of the same family of assembly or rebuilt
between product families. The research is part of an
ongoing EU-founded aircraft industry project
ADFAST*.
INTRODUCTION
Since aircrafts started to be built in aluminum, dedicated
tooling has been used for holding aircraft parts in
position during assembly. These tools are designed to
be dedicated only one assembly. For another assembly,
a new dedicated tool must be developed. The lead-time
of designing and building dedicated tools is long and
costly. The features locating and holding work pieces in
the tool are static in one configuration and lacks
flexibility, which tends to be a problem. Aircraft
manufacturers today are more and more becoming
specialists in a field to build a particular part of the
aircraft. They build the same airframe parts in different
models, sizes and production rates. In these cases
fewer tools would be a cheaper way to become more
competitive. Other areas of exploitation are the new
generation of aircrafts (i.e. unmanned aircraft vehicles)
where demonstrators are built in one-batch volumes.
Having one conventional tool for each airframe in this
case would be very expensive. This scenario demands
fewer assembly tools to be cost-effective. The fact that
Unmanned Aircraft Vehicles (UAV) is a new technology,
constant development of new systems will become even
more demanding on the production system in the future
to be able to adapt for changes. UAV is also an example
of a new trend to build demonstrators, i.e. one-batch
2002-01-2645
Henrik Kihlman
Department of Mechanical Engineering
Linkping University
Magnus Engstrm
Commercial Programs
Saab AB
production. If the production system could be flexible
enough, one-batch tooling systems could be converted
in to large-batch scenarios to handle ramp up production
in short notice. The question is: will it be possible to
continue using Dedicated Tooling in the future and still
be competitive?
There have been approaches to cope with changes in
products. One solution is Modular Tooling. This solution
uses standardized profiles and couplings that are
screwed together, so that the parts in the jig can easily
be adjusted. This solution can handle changes in the
centimeter range. Modular Tooling is one step to
increase flexibility in the fixtures.
Another approach to increase flexibility in the tooling is
CNC-controlled fixtures, more known as pogo-stick
fixturing where a bed of nails is used to change
configuration for new products. This solution also
represents the first step toward reconfigurability. CNC-
controlled fixtures are today on the other hand
expensive and complex.
Hence, using a standard robot for doing the
reconfiguration task enables the ability to reach a higher
level of flexibility and doing so at lower cost. This paper
is presenting a new tooling concept partly developed in a
EU-founded project ADFAST, called Affordable
Reconfigurable Tooling (ART) [1]. The ART-concept
consists of a static framework, dynamic modules, locking
mechanisms, manipulator interface and pick-ups. The
dynamic modules can be reconfigured by a robot
between different types of assemblies of the same
family. The static framework can due to modularity be
rebuilt for another family of assemblies. Reconfiguring is
done by guiding the robot with a laser tracker to dock
with a dynamic module. When docked the robot changes
the configuration of the dynamic module to a new
position. The laser helps the robot to compensate for
lack of positional accuracy and deflections. The concept
is based on having the robot not only to reconfigure the
dynamic modules but also for drilling and riveting,
therefore the term Affordable.
The paper starts out with a brief survey of the
conventional way of building aircraft assembly tools
followed by discussing existing solutions of tooling that
has flexibility. Finally the paper presents the general
idea of the ART-concept and the modules included.
CONVENTIONAL TOOLING
Production tooling in aircraft assembly is divided in two
categories. The first one is jigs. Jigs are workholders
designed to hold, locate and support a workpiece while
guiding the cutting tool throughout the cutting cycle [2].
The other one is fixtures, who are workholders which are
designed to hold, locate and support workpieces during
the machining cycle [2], in this case holding the
workpieces during assembly i.e. drilling followed by
riveting. Wang [3], defined fixtures as consisting of
clamps and locators, that must be capable of
positioning, holding and supporting a workpiece during
machining, assembly and inspection
Boring and drill jigs are something that aircraft producers
want to eliminate nowadays, and instead focusing on
automation using robotics or using CNC predrilled holes.
Sometimes it is difficult to tell the difference between jigs
and fixtures. Conventional Tooling is a comprehensive
term for how aircraft assembly fixtures has been, and
still are designed.
Conventional tooling is characterized by that they are
tailor made for each assembly. If a new product is
developed a completely new conventional tool must be
designed. This is why they sometimes are called
dedicated tooling [4] or hard tooling [5]. Conventional
tooling however is well known and they have been
practiced since aircraft started to be built in aluminum.
Pins and clamps are used to locate and hold airframe
parts in position before these are drilled and fastened
into one unit [6]. The pins and clamps are often
positioned on pick-ups. The pick-ups are elements
between the aircraft structure and the frame of the jig. In
extreme, the design of the pick-ups is generally unique
for each assembly. For the pick-ups there are rarely any
standards for the designer to use.
Designing new elements for each new assembly is
expensive and time consuming. Although conventional
tooling is safe and well known and almost anything can
be done due to its tailored design, the time and money
needed are two elements the aircraft companies do not
have a lot of nowadays, as the competition between
aircraft industries has increased. Earlier, aircraft
manufacturers made their own complete aircraft, and
now they are becoming specialist in a specific field of the
airframe. One example of a field is wing structures.
Although computer enabling digital mockups and
advanced machines with high speed machining
capabilities has become general practice, the
manufacturing and development lead-time for tool parts
still is too long and must be shortened.
The conventional way of building assembly tooling in
aircraft assembly is to use steel material, which is rigid
and has good thermal expansion properties. Steel is
also cheap and has good welding characteristics.
Aluminum is also utilized in conventional tooling,
particularly for movable fixture parts and turning fixtures
where low weight is preferable. In a typical development
of a conventional tool, the tool is designed with input
from the geometrical data of an aircraft structure that is
to be assembled. Although this data is needed in order
to understand what the fixture must look like, the
designer must start the designing of the tool before this
geometrical data is available. This must be done in order
to shorten the lead-time for the tool parts, which
generally is long. The parts with the longest lead times
must be sent to the manufacturer before the tool
designer gets the final geometrical data from the aircraft
designer. This is one of the reasons why there are a lot
of changes involved in the development of conventional
aircraft assembly tools.
To stay as a competitive aircraft manufacturer in the
future the tools must be standardized and flexible. The
assembly system must also be able to handle different
product types within the same system. This can only be
done if the flexibility and modularity is increased
throughout the assembly system.
FLEXIBILITY AS A SOLUTION
Flexibility in the production area and flexible production
systems has been spoken of since the beginning of the
1980. One clear reason for that is that most
manufacturing industries more or less have experienced
rising number of product variants and decreased product
lifespan. Nowadays, more different units need to be
constructed in smaller quantities, which increase the
demands on the production systems. The system is
supposed to manage more product variants and the
ability to change product generation frequently.
In the Visionary Manufacturing Challenges for 2020 [7],
moving from the current status of manufacturing to
manufacturing in 2020 there are six grand challenges
that would make the vision possible. One of them is to
make the manufacturing enterprises reconfigurable to
rapidly response to changing needs and opportunities.
This implies that the manufacturing systems must be
adaptable and reconfigurable and have a flexible
forming process so that no hard tooling is needed.
The problem with companies using product- or variant
specific tools is that the cost needs to be distributed
among fewer products, which gives heavily rising
product costs as a result. A report from Tichem et al [8],
defines flexibility as the capability to react to changes in
the environment. The same report discusses the
problem with automation in the assembly process; it
often leads to the development of case-specific
solutions. It results in high degree of assembly
engineering and high capital investments for each
product to be assembled because of product-specific
equipment that has to be developed and used. This
phenomenon does not permit the application of
assembly automation for small series production.
Flexible fixturing involves employing a single fixturing
system to hold work-pieces of various shapes and sizes
[9]. A new paradigm was presented by Mehrabi et al.
[10], called reconfigurable manufacturing system (RMS),
which they defined as the next step after Flexible
Manufacturing Systems (FMS). They stated that RMS is
designed for rapid adjustment of production capacity and
functionality, in response to new circumstances, by
rearrangement or change of its components. Applying
RMS will go beyond mass, lean and flexible
manufacturing towards reduction of lead-time for
launching new manufacturing systems and also enable
rapid upgrading and quick integration of new process
technologies and functionalities into existing systems.
A product family is defined as one or more part types
with similar dimension, geometric features, and
tolerances, such that they can be produced on the
same, or similar production equipment [10],[5]. In figure
1, the three most common product families in aircraft
production are presented.
Further in this paper three levels of flexibility is going to
be presented. The first level of flexibility represents the
ability to adjust a fixture for minor changes within one
product. The second level of flexibility represents the
ability to reconfigure the fixture between parts of the
same product family. The third level of flexibility is the
ability to be able to rebuild the fixture between product
families.
CURRENT SOLUTIONS
Flexible Fixturing is not new. There exist several ideas to
increase the amount of flexibility. The first approach is
Modular Tooling. Flexibility in this paper is defined as
how quick the system can be reconfigured on a macro
scale. Macro scale is the ability to build vehicles as cars,
aircrafts etc.
MODULAR TOOLING
Modular Tooling represents the first level of flexibility in
assembly tooling. The idea is based on building the tool
by a standardized kit of parts and couplings. Today there
exist several suppliers for Modular Tooling who have a
wide range of parts and can ensure a certain quality.
Modular Tooling has been used in the car industry for a
long time and is recently being introduced in aircraft
industry.

Figure 1: Three different families of structures. Skin

structures to the left, wing boxes in the middle and
aircraft fuselages to the right.

If one production system where to be used for the

handling of all three families of structures, a greater
need of geometrical flexibility would be needed, and that
is what this paper is trying to do. It is about extending
the concept of flexibility not only to being able to handle
Figure 2: A Modular Tool with an adjustable pick-up.
changes in within one product, or even to be able to
reconfigure the fixture between products within one
To build these tools, extruded aluminum profiles are the
product family. We also try to present an idea to reach
most commonly used material. The profiles are screwed
flexibility to being able to reconfigure an assembly tool
together with couplings, which also can be bought by the
between product families.
supplier of the profiles. Modular Tooling uses pick-ups,
which compared to the pick-ups in conventional tooling,
Maybe one of the structures is possible to use a system
are adjustable. An interesting property with the
that could be reconfigured between different structure
adjustable pick-ups is that the accuracy in the frames
types within the same family. If however one would like
where the pick-ups are attached does not have to be
to change the configuration of the tool into another
high. After the pick-up is roughly attached to the frame, it
product family, the tool would definitely have to be
can be adjusted by loosening some screws in the
configured in a wider range. This is where the ability to
adjustment points starting with the ones closest to the
rebuild comes in. Rebuilding is when the dynamic units
frames, see figure 2. As changes in aircraft design are
of the tool can be dismantled from one position and
common, changes in the tooling will also become
mantled in a new position, which is far beyond the work
necessary. The adjustable pick-ups make it possible to
volume of the reconfigurable tool. After the rebuilding is
handle these changes.
done the tool can look entirely different.
The adjustable pick-ups in Modular Tooling have proven
to reduce lead-time in the development and building of
the tools. This can be done because the position of the
frames of the tool is not so demanding, as the pick-ups
are adjusted to the final position of the locators or
clamps. By not having to put a lot of effort into
positioning the profiles, the time to build can be
shortened. Reduced lead-time in Modular Tooling is also
a result from using libraries of standard parts when
designing the tools. When designing the Modular Tool,
the designer can pick parts from a library instead of
having to design new features for every new tool.
Modular Tooling is a good example of a technology
where the system can handle changes within the same
product type, i.e. the first level of flexibility. The fixture
can change in the range of centimeters; therefore it is
not possible to do any reconfiguration to assemble
different types of structures with one tool configuration.
However, the tool can be dismantled and rebuilt to
assemble another type of structure. But this is rather an
example of reusability rather than an example of
reconfigurability. Rebuilding the tool would take a couple
of weeks.
FLEXIBLE TOOLING
Flexible tooling is the first step towards increased
flexibility. Flexible tooling is able to reconfigure the
movable fixture units sequentially one by one until the
entire tool is reconfigured. The can be reconfigured for a
new product within the same product family.
Reconfiguration of all fixture units is possible to do within
the hour.
A lot of research was done in flexible tooling back in mid
80th and early 90th. One of them was at McDonnell
Douglas, Long Beach, CA, where they developed an
automated flexible assembly cell to gain flexibility and
reduce tooling costs [5]. They replaced the separate
tools with a single, flexible tool that serve a family of
assemblies and provided automated inspection. The cell
automatically assembles, rivets and inspects small, flat
bulkheads. They used a gantry robot to configure the
flexible jig and loaded detail parts into the jig. Goodman
claimed to replace 250 conventional jigs with one flexible
tool.
Manufacturing of sheet-metal parts is considered difficult
to automate because of the complex geometries
involved. This problem was solved by M.I.T. in mid 80th
by Youcef-Toumi et al. [12],[13], Youcef-Toumi and
Buitrago, [14], Fields et al. [15], Asada and Fields, [15],
and Asada and By, [16]. They used a method called
Robot-Operated Adaptable Fixture to change the
configuration of a fixture bed by manipulating vertical
supports (fixture elements). Changing the length and
position of the vertical supports different sheet metal
parts could be machined, i.e. drilled. They solved the
problem to automate the drilling of sheet-metal parts i.e.
skin structures.
Another approach on flexible tooling manipulated by a
robot, was presented by Engstrm [17]. Engstrm
strived for elimination of specific jigs for assembly by
defining flexible jigs adapted for automation. A new
concept was defined and patented [18]. A defined 3-axis
rail system was manipulated by a robot and made
reconfiguration of the tool possible. The system was
able to handle skin structures. Simulations of the
concept were done, but no physical demonstrator was
built.
CNC-CONTROLLED TOOLING
Increasing the level of flexibility to the third level of
flexibility move us to CNC-controlled tooling or hyper-
flexible tooling. CNC-controlled tooling has the ability to
configure the fixture for different structures similar as
Flexible Tooling, however all moveable fixture units can
be configure simultaneously. Goodman [5], suggested
this as the next step for flexible assembly, where the
flexible jigs position themselves using multiple-stepper
motors. One complete reconfiguration of the tool is
possible to do within a couple of minutes.
A similar project of this research was done in the FAS
(Flexible Assembly Subsystem) program presented by
Wright and Sarh, [19], and Olsen [4]. They had the
scope to increasing quality while reducing costs. They
developed and demonstrated flexible and automated
workcells that could replace dedicated tooling and be
able to quickly reconfigure itself for new types of
subassemblies. A robot was used for assembling the
aircraft parts and the flexible fixture consisted of four
three armed robots with a total of 48 axes controlled by
the fixture controller. Where each of the 12 arms has a
ball-joint mounted between the last link and the end-
effector quick-change unit. The fixture controller can
either release or freeze the ball-joint unit. Its end
effectors are either clamps or suction cups mounted on
quick changers.
A more recent approach to CNC-controlled tooling was
presented by Yingjie et al. [20]. Yingjie had an approach
on a flexible workcell for aircraft wing spar assembly.
They developed a concept of a fixture with 76 step
motors controlled by a central controller. The concept
was suitable both for different type of aircrafts. and also
for the left wing and the right wing of the same aircraft.
This work with CNC-controlled tooling has been carried
on by Kostyrka, [21], and MTORRES [22]. They use
pogo sticks, which consists of a cylinder, which can be
extracted or reversed in one axis. A clamping device is
used to lock the cylinder. The sticks also have servo
engines and encoders to locate themselves to the right
configuration. The sticks are put together into a matrix,
forming a bed of nails, se figure 3.

Figure 3: A bed of nails
CNC-controlled tooling is mostly used to fix aircraft skins
for drilling, or for milling the edges. This solution is
mostly used as a holding device for machining
operations. No advanced assembly is necessary. On the
top of each pogo stick it is common to have a vacuum
cup, which sucks the part in position. The system can
change in the range of decimeters, and makes
reconfiguration possible, but in a rather restricted
manner. CNC-controlled tooling manages to reconfigure
within a product family.
A presentation of three different main categories of
flexible fixture system where presented by Thompson,
[23], and Thompson & Grandi [24]. The first one was
Modular Fixture Kits, which is an extension of fixture
building using standard work holding devices. The next
one was a Phase change fixture that forms a supporting
structure from an enveloping matrix. The component
machined is removed by re-fluidizing the matrix, which
can be recycled. The third category was Programmable
fixtures, which could be reconfigured to accept different
components.
Figure 4 shows the summary of existing tooling
technologies with a hyperbolic curve. Conventional
tooling is not practical to reconfigure at all. Modular
tooling take a weeks to build/rebuild, flexible tooling is
done within the hour and CNC-controlled tooling/hyper-
flexible tooling takes a couple of minutes.
Figure 4: The summary of current tooling solutions
A NEW TOOLING CONCEPT
Using ideas from Conventional Tooling, Modular
Tooling, Flexible Tooling and CNC-controlled a new
concept is being developed Affordable Reconfigurable
Tooling, see figure 5. The concept is based on having a
robot to handle drilling, riveting and the reconfiguration
movement, similar as in flexible tooling. The tooling work
volume is surrounded by a static frame work that are
modular such as Modular Tooling. In the static
framework there are dynamic modules of different
configurations, one very similar as in the pogo-stick
mentioned earlier. The robot can reconfigure the
dynamic units. The static and the dynamic system
together form a universal tooling system. To handle the
large range of products on the end of the dynamic
modules pick-ups can be attached, which is basically
based on the same idea as Conventional Tooling.
Everything in the tooling system is modular and
exchangeable with ART-system configurations.
Figure 5: An overview over the ART-concept
Further in this paper each module of the ART-system
will be described more in detail.
STATIC FRAMEWORK
The static framework is the fundament that the
dynamical system is attached to. It consists of beams
that are rigidly connected to the floor. The static
framework defines the work envelope for the tool.
Depending on which product family that is to be
assembled the static framework is configured differently.
Changes to the static framework are called rebuild and
is preferable done as rare as possible, whereas the
rebuild process takes much more time than the
reconfiguring process. But the rebuild enables in theory
a indefinite change-over range.
The static framework is divided into two different
categories. The first one is the welded steel framework,
which is designed similar as the welded steel framework
in conventional tooling. This solution is configured only
once, whereby changes are not possible to do at all. The
second solution is the modularized framework method,
which is based on the same idea as modular tooling.
This solution is base on having the beams attached
together using screws, whereby rebuild possible to
perform.

Both the welded steel- and the modular framework can

use two separate methods of connecting the dynamic
system. The first method is based on having a
coordinate hole pattern were the dynamic units are
attached through quick couplings or screws. In this case
each dynamic modules can be positioned in a discrete
pattern. The second method is based on having rails for
the dynamic units to travel continuously on the static
beams.

DYNAMIC MODULES

The dynamic framework is mounted on the static

framework. It can be moved in order to reconfigure the
position of a pick-up. The movement (longitudinal and/or
rotational) can be done continuously and/or in defined
steps along a number of axis. The dynamic framework
includes transmission devices with bearings to allow
relative movement (play free) between different
components in the dynamic framework. The dynamic
modules is divided into five different modules:
The 1-axis conventional pogo stick
Figure 7: The cannon concept
Figure 7 shows the cannon concept that is basically the
same as the pogo stick, with the difference that it has an
extra rotational axis. The pogo stick has an interface
where the cannon feature can be attached. Sometimes it
is difficult to get access to pick-up points from only one
angel, and then the cannon module can come in hand.
This unit also has the advantage to be able to stand
freely on the floor.

The 3-axis rail module

Figure 6: The pogo stick concept

Figure 6 shows the 1-axis conventional pogo stick

concept, which is based on using conventional pogo
sticks. The pogo sticks are off-the-shelf products that
today are commonly used in bed-of-nails presented
earlier. The robot docks to the module and changes its
configuration by extracting/retracting the length of it.
When the new position is reached, the module activates
its looking mechanism and it stays locked until next
reconfiguration. The advantage with this module is its
simplicity. Very often no more than a 1-axis movement is
necessary to pick up a point in space. Due to its
simplicity it is also the cheapest solution of all dynamic
modules.
The 2-axis cannon module
Figure 8: The 3-axis rail concept
Figure 8 shows the 3-axis rail concept is also built on the
pogo stick principal accept that it has a 2-axis linear
movement in the plane. The movement in the plane is
enabled through a rail system that gives a continuous
movement of a column of pogo sticks. If a column
contains more than one pogo-stick it will not be a fully a
3-axis system because of the fact that the pogo sticks in
that column will rely on each other. They will not be
possible to move independently. The reconfiguring
sequence is the same as for the 1-axis pogo, except for
the movement in the plane, which also is achieved by
the robot.

The 6-axis hexapod module

Figure 10: The cradle concept

Figure 9: The hexapod concept
Figure 9 shows the hexapod solution that is based on
having six links connected in parallel between the top-
and the bottom plate. This is based on ideas from
parallel-kinematic machines where the parallel links
gives enhanced stiffness. But the hexapods presented in
the ART-concept have no actuators and, therefore it is
called a parallel-mechanic unit. Askbrink [28], made an
analysis of the optimal leg configuration for a parallel-
mechanical device designed from aspects on the
requirement in aircraft assembly. The hexapods can
move freely in six degree of freedom and is very stiff
when locked. This solution is the most advanced
solution and also the most expensive. It is used when
there are requirements on handling high loads or hold
parts with great stiffness.
Figure 10 shows the cradle module that is a very flexible
dynamic unit. It can move freely in six degree of freedom
between the set-ups. The foundation of the unit is a 1-
axis pogo stick, which not only has one extra rotation as
in the cannon module but two. The final three degrees of
freedom is achieved from a ball joint in the end of the
rod. Thysell and Denholt [25], made a first approach on
this concept.
Interestingly the position accuracy of the hexapod- and
the cradle solution is not given by the accuracy of the
static framework itself but by the manipulator that moves
the framework.

The 6-axis cradle module

LOCKING MECHANISM
The locking device is a module that is integrated in the
transmissions and bearings in the dynamic modules.
The locking device has the functionality to lock the
different components in the dynamic modules after the
robot has configured them. It is important that the pick-
up position does not move during the locking sequence
and that the locking force is high enough so the position
does not change throughout the assembly process.
Important is also the safety of personal working in the
fixture. This is a risk when working with high pressure
hydraulic. Using a hydro-mechanical locking principle
from ETP Transmission AB has solved this problem.
Figure 11: The ETP-locking mechanism
As shown in figure 11, the principle is based on only
pressurizing the locking mechanism during the locking-
or unlocking sequence. When the system is pressurized
it moves a wedge that is self-locking, whereas the
friction between the wedge and the housing prevent the
wedge from moving without hydraulic pressure. Although
hydraulic pressure is needed to move the wedge
between locked and unlocked mode, the pressure is not
required to withhold the locking force.
MANIPULATOR INTERFACE
In the demonstrator currently being built at Linkping
University the tool changing system Capto from
SANDVIK has been chosen as the interface between the
dynamic modules and the robot. Figure 12 shows a
robot docked to a dynamic module using the Capto
system. It allows the robot to configure the dynamic
modules accurately through laser guidance hence it is
Important that the docking and undocking sequence
does not effect their position. The principle is based on
using a chuck on the TCP of the robot. On each dynamic
module, drilling machine, riveting machine or anything
that the robot interacts with has a male part of the Capto
system. This tool changer system has good repetitivity
(2m) and can handle loads up to 1200 kg. The release
movement gives a small movement on about 0.3mm,
which is small enough not to permanently move the
dynamic modules after releasing the grip of the male
part of the dynamic modules.
PICK-UPS
The pick-up is mounted on the dynamic framework. It is
the device that creates the interface between ART and
the airframe structures. This means that the pick-up
connects the airframe structure to the universal tooling.
It can be dismounted from the dynamic framework and
exchanged to other pick-ups depending on access to
different location holes in the structures. In order to
handle many different structures in one system, the pick-
ups can be specially design, dedicated to each structure
to assemble. This is also on way to slowly move over
from a conventional system to the flexible system.
Conventional, dedicated pick-ups can be used the same
way as they used to. The dedicated pick-ups can on the
other hand be standardized, so for instance all clamps
and pinholes are the same. On each pick-up a small
manual Capto chuck is attached. This means that the
pick-up uses the same interface as the robot does for
reconfiguring the dynamic modules. The robot has
positioned each module with high precision, and it is
therefore an advantage to use that same datum point.
The Capto interface is also a convenient interface to use
due to its strength and simplicity. Both the interface and
the manual chucks are mass produced off-the-shelf
products and therefore cheap.
Figure 12: The robot docked with the parallel
mechanical unit through the Capto interface
THE METROLOGY SYSTEM
In the final concept the robot is to be integrated with a 6
DOF motion control in real-time from Leica [26],[27]. So
far in the project, experiments regarding robot control
supervised by a 3 DOF laser has been tested. The robot
is trusted to move by its own controller system to a final
position i.e. a pick-up. When the robot has reached the
position the laser measures the position of the robot
TCP. In general a good calibrated robot can move to a
position in space for about 0.5 mm. That is the positional
accuracy. The repetitive accuracy on the robot used in
the project was about 0.07 mm. The robot used was an
ABB robot, IRB4400 with s4c+ controller system. After
the robot has positioned a dynamic unit to around 0.5
mm, the laser corrects the robot down to around 0.1
mm, which is within the required accuracy for the
assembly. The method is similar to single laser-tracking
interferometry (ISO/TC 184/SC 2 N 293). Later in the
project, the robot will be integrated with the Leica
LTD800 [27], which measure the robot position and
orientation in 6 DOF.
COST SAVINGS
Cost savings can be done on many areas using ART.
Most of the parts of the tool modules can be
manufactured with low accuracy and is therefore cheap.
No additional controller system is needed for positional
the dynamic modules due to the use of a robot. Only one
active system is used. The robot is not only used for the
reconfiguration of the tool but also for the automation of
drilling, countersinking, riveting and material handling.
These tasks if not done manually would require separate
specialized machines, which tends to get expensive.
The tool system is built up from standard off-the-shelf
products. This not only saves money but also ensure
quality due to having a supplier that can guarantee
certain properties of their products. Only the amount of
flexibility that is needed can be chosen, where the less
flexible modules are cheaper than the more flexible
ones. Quick changeover time between set-ups results in
less lead time. Finally changes in the products are easy
to adapt for.
FUTURE RESEARCH
Although this project has undergone several integration
tests with the robot and other systems such as laser and
fixtures, most of the concept is so far conceptual. The
remaining time in the project all the enabling
technologies that are needed for the concept to work will
be tested and analyzed. This will map which areas
needs further development before an industrial
implementation. The approach to that will be to build a
full-scale demonstrator at Linkping University whereby
each module in the system will be physically evaluated.
One major task will be to solve the problem of positional
accuracy in robots. Today robots have good
repeatability, but bad absolute accuracy. This will require
some additional measuring loop for the robot. Lasers,
cameras or other systems can solve this. However,
lasers are expensive and cameras still need a lot of
computer compiling. This system integration will require
a lot of research effort.
From further analyzing future scenarios the question on
which level of flexibility that is required will be answered.
This answer will be answered from two aspects: how
large a change-over range do we need, and secondly
how quickly do we want to reconfigure?
CONCLUSION
In this paper a reconfigurable tool concept has been
presented. It is a universal tool for aircraft assembly that
can be adjusted to adopt for smaller changes,
reconfigured between products and rebuild between
product families. This is done by using a standard
industrial serial linked robot to reconfigure the tool.
Using a robotic device to reconfigure the tool, to drill, or
to rivet etc the absolute accuracy in robots must be ten
times better than today. This will be done by external
metrology system. One of the biggest conclusions from
the project is that in order to be a competitive aircraft in
the future flexible and modular assembly tools will be
needed.
ACKNOWLEDGMENTS
This work is part of the EU-founded project ADFAST
(Automation for Drilling, Fastening, Assembly, Systems
Integration, and Tooling). Special thanks go to project
manager John Andersson. The authors would also like
to thank SANDVIK and Leica.
REFERENCES
1. Kihlman, H., Reconfigurable Tooling for Airframe
Assembly A state-of-the-art Review of the Related
Literature and a Short Presentation of a new Tooling
Concept, CIRP 1st International Conference on
Reconfigurable Tooling, May 21-22, 2001
2. Hoffman, E.G., Fundamentals of Tool Design,
Society of Manufacturing Engineers, Dearborn,
Mich., 1984
3. Wang, M. Y., Characterizations of positioning
accuracy of deterministic localization of fixtures, in
2002 IEEE International Conference on Robotics
and Automation, Washington DC, May 2002
4. Olsen, Howard B., 1990, A Flexible Robotic Work
Cell for The Assembly, IEEE International
Conference on Robotics and Automation May 13-18
1990
5. Goodman, H., 1990, Assembly cell speed riveting,
Tooling & Production, Dec., 1990
6. Trappey, J. C. and Liu, C. R., A literature survey of
fixture-design automation, International Journal of
Advanced Manufacturing Technology v 5 n 3 Aug
1990 p 240-255 0268-3768 In English 48 Refs
EI90120082366
7. National Academy Press Washington, D.C.1998,
Website visited 2001-10-25, Visionary
Manufacturing Challenges for 2020,
http://www.nap.edu/readingroom/books/visionary/
8. Tichem, I.M., Storm, I.T., Vos, I.J., 1999, How to
achieve a breakthrough in industrialisation of flexible
assembly automation, Proceeding of the Ninth
International FAIM Conference, June 21-25, 1999
9. Shirinzadeh, B., "Issues in the design of the
reconfigurable fixture modules for robotic assembly",
Journal of Manufacturing Systems, 12(1)(1993)1-14
10. Mehrabi, M.G., Ulsoy, A.G, Koren Y., 1998,
Reconfigurable Manufacturing Systems: Key to
Future Manufacturing, Japan and USA Symposium
on Flexible Automation, July 12-15, 1998
11. Youcef-Toumi, K., Bausch, J. J., and Blacker, S. J.,
Automated setup and reconfiguration for modular
fixturing, Proceedings, Symposium on
Manufacturing Systems - Design, Integration and
Control, Manufacturing Conference, Atlanta, Ga., pp.
121-128, 1988
12. Youcef-Toumi, K., Liu W. S. and Asada H,
Computer-Aided Analysis of Reconfigurable
Fixtures and Sheet Metal Parts for Robotic Drilling,
Robots & computer Integrated Manufacturing, Vol. 4
No. 3/4, pp. 387-393
13. Youcef-Toumi, K., and Buitrago, J. H., Design of
robot-operated adaptable fixtures, Proceedings,
Symposium on Manufacturing Systems - Design,
Integration and Control, Manufacturing International
Conference, Atlanta, Ga., pp. 113-119, 1988
14. Fields, K. Youcef-Toumi and H. Asada. "Flexible
fixturing and automatic drilling of sheet metal parts
using a robot manipulator ",
15. H. Asada and A. Fields, "Design of flexible fixtures
reconfigured by robot manipulators ", Proceedings,
Symposium on Robotics and Automation, ASME
Winter Annual Meeting, Miami Beach, Fla., pp 251-
257, 1985
16. Asada, H. and By, A. B., Implementing Automatic
Setup Change via Robots to achieve Adaptable
Assembly, 1984 American Control Conference, San
Diego, CA
17. Engstrm, M, 1998, Flexible Workshop for Airframe
Assembly, Nouvelle Revue Deronautique et
Dstronautique, Nr. 2, 1998, 3rd Aero Days Post
Conference Proceedings
18. Engstrm, M., Verktyg fr fixering av skrovdetaljer, Production System
Patent publ. nr., WO, 0179061
19. Wright, K., and Sar, B., 1989, Computer integrated
Quality Assurance for Robotic Workcells in
Aerospace Manufacturing, Robots in Aerospace
Manufacturing, SME, February, 1989
20. Yingjie Li, Behnam Bahr and Xiaobin Chen, The
design of a Flexible Fixture & Workcell for Aircraft
Assembly, CAD/CAM Robotics and Factories of the
Future, Middelsex University, London, England,
August 11-13, 1996
21. Kostyrka, P.A, Kowalsky, J, Flexible active and
passive pogo fixturing systems for aircraft and
aerospace applications, SAE Aerospace Automated
Fastening Conference & Exposition, 2000
22. MTORRES Group, Website visited 2002-06-27,
http://www.mtorres.com
23. Thompson B.S. 1984 Flexible Fixturing A Current
Frontier in the Evolution of Flexible Manufacturing
Cells, ASME Paper No. 64-WA/Prod-16, The
American Society of Mechanical Engineer
24. Thompson B.S. & Grandi MV, 1986, A Commentary
on Flexible Fixturing, Applied Mechanics Review,
19(10): pp. 1365-1369
25. Thysell and Denholt, 2002, Reconfigurable Tooling
for Aircraft Assembly, Final thesis report, january,
2002 Thysell and Denholt, 2002, Reconfigurable
Tooling for Aircraft Assembly, Final thesis report,
january, 2002
26. Dold, .J, Loser, R., Machine Calibration and
Guidance with Laser Tracker, SAE Aerospace
Automated Fastening Conference & Exposition,
2002
27. Leica Geosystems press release, Take 3D Portable
Metrology to new Dimensions A new generation of
Leica Trackers has arrived, 6 June, 2002
28. Askbrink, 2002, Optimisation of Parallell Mechanical
and Adjustable Assembly-jig, Final thesis report,
january, 2002
CONTACT
Henrik Kihlman
Linkping University
Phone (+46) 13-288 974
E-mail: Henrik.Kihlman@ikp.liu.se
Magnus Engstrm
Commercial Programs
Saab AB
Phone (+46) 8-673 5439
E-mail: Magnus.Engstrom@saab.se
ADFAST:
www.euadfast.com
 Appended Publication IV

Kihlman, H., Eriksson, I. and Ennis, M., (2002) "Robotic Orbital Drilling of Structures
for Aerospace Applications ", SAE Aerospace Automated Fastening Conference &
Exposition, October. 1-3

Robotic Orbital Drilling of Structures for Aerospace
Copyright 2001 Society of Automotive Engineers, Inc.
ABSTRACT
This paper describes ongoing research into orbital
drilling using standard industrial robots. The research is
a part of an ongoing EU funded aircraft industry project
ADFAST*. Generally it is difficult to use standard
industrial robots to automate drilling in the aerospace
industry. The stiffness of the standard robotic device is
not sufficient to resist the deflections caused by the
cutting forces from the drilling process, therefore it is
difficult to achieve the tight hole tolerance requirements.
Orbital drilling creates lower axial cutting forces
compared to conventional drilling and therefore allows
the use of low-cost standard industrial robots for drilling
holes within the required hole tolerances. This paper
presents results from a study where forces, moments
and dislocations produced during orbital- and
conventional drilling have been measured.
INTRODUCTION
The process of drilling holes in aircraft assembly is far
from being fully automated. The cost for the assembly of
an aircraft consumes up to 40% of the overall
manufacturing cost [1]. In order to automate aircraft
assembly manual drilling must be replaced by robotic
drilling. Efforts to use standard industrial robots (IRB)
have been tested because they have high flexibility [2].
However large flexibility reduces the stiffness due to
serial linked arms. This is one of the reasons why other
solutions like parallel kinematical machines have been
tried as a solution [3], or by using CNC-machines.
Tripods on the market today cost over three times more
than conventional industrial robots and CNC machines
up till ten times the amount of a conventional IRB. CNC-
machines also require a lot of floor space. Aircraft
structures are often complex and a device of less size is
necessary for access. Conventional IRBs have although
been used, but often together with drilling templates
which requires a lot of dedicated costs and space. The
forces from the drilling process must be lower if IRBs are
2002-01-2636
Applications
Henrik Kihlman
Department of Mechanical Engineering
Linkping University
Ingvar Eriksson and Mark Ennis
Novator AB
to be used without drilling templates. Orbital drilling is a
drilling method that generates much less cutting force
compared to conventional drilling. This paper
investigates the potential of using an orbital drilling
device in order to achieve superior hole quality while
drilling with a serial linked robot. Several tests were
made where force, deflection and hole quality was
measured both in stiff rigging and with a robot.
ROBOT DRILLING TODAY
DRILLING TEMPLATES
In manual drilling templates are necessary to give the
cutting tool the right position and orientation. Automated
drilling with robots has also been done using drilling
templates [4]. All manufactured parts have a
manufacturing tolerance. This variation affects their
placement within the assembly fixture. The drills are
guided by the template which is connected to the fixture.
Therefore the hole positioning and countersunk holes
will be out of round and either too deep or too shallow
proportionally [1].
LARGE DRILLING MACHINES
Large drilling stations are ten times more expensive than
a conventional industrial robot and are often dedicated in
their design. Whinnem and Nystrm [3], stated that the
attempts to automate the assembly of todays aircraft
structures have focused towards very large and rigid
support structures that fit inside the working envelope of
even larger rigid numerically controlled machine tools
where drilling and fastener installation tasks are
performed. One example of a highly automated machine
for automation of wing panel assembly was presented
by Hartman [5].
Airplane construction is generally characterized by low
volume batch type production, and since a typical
airplane requires many different sheet metal parts, each
requires a unique fixture.
ROBOT ACCURACY
A problem with robotic drilling is the inaccuracy of the
positional accuracy. A robot is always specified by the
supplier in terms of repetitive accuracy. A high
performance anthropomorphic robot today has a
repetitive accuracy down to 0.07mm. That is enough for
general aircraft assembly requirements which require
the drill to be positioned within 0.01-0.02mm [6],[7].
Repetitive accuracy is the robots ability to return to a
position in space where it has been once before. This
was well suited in the old days when most of the
programming of the robot was done by teach-in,
whereby the operator eliminates all error by eye [3].
Today, robots are offline-programmed in robot
simulation software. This method requires the robot to
find a given position in space within a specified
tolerance. When a robot is given a coordinate to where it
has never been before errors in the robot hardware and
software are accumulated to a value called positional
accuracy or absolute accuracy. The positional accuracy
in a good quality robot today is seldom better than 1mm.
In a level 1 calibrated robot accuracy can be reduced to
0.5mm [8]. One must also have in mind that this
tolerance is achieved in an unloaded robot, and in a
restricted area of the robot work volume. Although the
software in todays robots can compensate for loads it is
only for gravitational loads. When a robot is used for
holding a drilling device during drilling it will very often be
in the horizontal direction, which the robot cannot
compensate for.
ROBOT BASE-STIFFNESS
Many different ways have been tried to compensate for
dislocations in robots today [9],[10], i.e. where an
external measuring loop is used to improve the final
position of the robot TCP (Tool Center Point). These
methods are basically about moving the robot to a
position in space and when the position is reached the
measurement system can verify and correct the position.
Gooch [9], stated that metrology is an enabling
technology and applications are being extended from
inspection to control of the manufacturing process itself.
A state-of-the art laser measures down to 50m.
However, a state-of-the art IRB has an incremental
accuracy of 70m, which limits the positional accuracy.
Even if a high enough positional accuracy could be
achived in the robot; the ability to adjust the robot to this
is limited in time. Every time a correction of positional
accuracy is carried out the measuring equipment sends
data to the robot. This is called synchronization. Hence,
even if a laser can measure the position of the robot
TCP in 1000 Hz, this does not mean that the robot can
be corrected at that speed. This is due to the limitations
in the interfaces in industrial robots today. They are not
well prepared for using external measuring equipment to
improve positional accuracy at high speed. The
interfaces to a robot controller today are floppy disks,
serial link or Ethernet. One must be aware of that it is
not the baud rate of a connection that defines the speed.
It is rather the ability to send a data package and verify
that it reached its destination that limits the
synchronization speed between equipment on a
network.
Hole drilling introduces rapid changes in forces. This
means that a measuring system although having a high
internal speed will not be able to compensate for the
robot deflection during a drilling cycle. Several other
approaches have been tested besides using drilling
templates. One straightforward solution is to let the robot
push against the sheet metal to be drilled. This fixates
the robot end-effector against the plate, which eliminates
certain movements during drilling. This is however a
rather cumbersome operation. When a robot is exposed
to a force at the TCP, it will not stay orthogonal to the
workpiece anymore [2]. It will change the TCP in 6
degree of freedom. A way to solve this problem is to add
an additional interface between the TCP and the end-
effector that is flexible. This will enable the drilling
machine to align itself vertical to the workpiece, whereby
the rotational deflections in the TCP are compensated.
Another way presented in this paper is to use a drilling
process where the forces are lower. Theoretically the
robot will deflect no matter the force exerted. However,
the robot has some base-stiffness which means that if
the deflection caused by a force is small enough it can
be neglected. A lot of problems would be solved if the
forces from the drilling process where so small that the
robot base-stiffness would be enough to avoid
deflection. This paper presents orbital drilling, which is a
drilling method that creates small forces.
ORBITAL DRILLING
Orbital drilling is based on moving the tool
simultaneously in both axial and radial directions by
feeding the tool through the material and causing the
tool to rotate about its own axis while at the same time
revolving the tool. Since the center of the tool orbits
about a stationary hole center the thrust force is
minimized and many of the problems associated with
traditional drilling (stationary tool center) are eliminated.
Moreover the tool diameter is less than the hole
diameter enabling efficient chip and heat extraction. The
cutting edge is partially and intermittently in contact with
the work piece allowing for efficient chip extraction. The
eccentricity is adjustable, thus the precision of the hole
is independent of the precision of tool. This in turn
makes it possible to drill high precision holes with low
precision tools and to drill holes of different diameters
with tool of one diameter. As seen in Figure 1, orbital
drilling can machine different kinds of axis-symmetrical
holes such as cylindrical, conical and complex shaped.

Figure 1: The principal of orbital drilling
The orbital drilling cutting tool has axial and radial cutting
edges (see Figure 2). One edge is for performing the
milling operation, which cuts in the radial direction. The
second one is on the face of the tool, which cuts in the
axial direction. A normal milling operation involves
moving the cutting tool in radial direction. Drilling is
obviously moving only in axial direction. Orbital drilling
moves in a helical movement through the material.
Figure 2: The orbital drilling tool
The only earlier publication on orbital drilling was made
by Lindquist et al [11]. It was presented as a study of
using orbital drilling as a means to eliminate
delamination of a carbon sandwich construction. It was
concluded that orbital drilling eliminated delaminations.
Figure 3: The portable orbital drilling unit
Todays orbital drilling system consists of two
configurations. One is the TWINSPINTM CNC, which is a
programmable orbital drilling unit. In Figure 3 the second
one is shown i.e. the TWINSPINTM Portable Orbital
Drilling Unit (P ODU). The P ODU, which was used for
the experiments in this paper, was a prototype drilling
motor intended for drilling high quality holes within the
tolerances typically found in aerospace applications.
The P ODU is a four-degree of movement mechanism. A
machining spindle rotates the cutting tool and a hydro-
pneumatic system with an adjustable feed rate is
configured for moving the cutting tool in an axial
direction. An orbital motor is configured for rotating the
cutting tool about a principal axis, and a radial offset
mechanism is configured for adjusting the radial
distance of the cutting tool axis from the principal axis.
EXPERIMENTS AND RESULTS
The experiments in this paper show that orbital drilling
could be an enabling technology of robotic drilling. This
was concluded by measuring and comparing cutting
forces in both orbital drilling and conventional drilling. A
similar study, but with the scope of conventional drilling
in composites has been made by Fernandes et al. [6].
Deflections caused by the cutting forces were measured
in the robot. This shows that it is preferable to use
something similar to a rubber ring to reduce the risk of
movement in the robot when the direction of the forces
in orbital drilling change continuously throughout the
drilling process. The method used to do this was to
perform the tests both for conventional- and orbital
drilling first in a reference case followed by similar tests
in the robot. All test were done in 6mm thick aircraft
aluminum 7475-0 series. Further in the paper feed rate
is defined as the traveling speed for the cutting tool
trough the material in axial direction.
FORCE MEASUREMENTS
The reference measurements were performed in a NC-
machine, where the machine was used for holding the
drilling machines and specimens rigidly. The drilling
machines were attached in the NC-machine chuck and
the specimens were attached in the NC-fixture, see
Figure 4, and 5. The orbital drilling prototype used for
these experiments was a prototype and therefore had
some restrictions. Firstly the spindle speed was constant
at 12000 rpm. Secondly the orbital rotation speed was
constant at 120 rpm. The parameters that were possible
to change were hole diameter and feed rate. Therefore
three different hole diameters and several different feed
rates were tested.
Figure 4: The P ODU attached to a NC-chuck
The conventional drilling machine was an Atlas Copco
with a self-feeding mechanism. It was set to work at
3000 rpm.
 The axial edge of
the cutting tool
fully exits material

The radial edge of

the cutting tool
Entering material
exits material

orb_05

Cutting edge has

full contact

Figure 7: Orbital drilling with feed rate: 0.7mm/s, 6

mm hole diameter and a 4 mm cutting tool.

The orbital drilling process had a sinusoidal force curve

Figure 5: The conventional drilling machine attached
to a NC-chuck
The force measurements were made with a 3-axis force
sensor bed, type Z15063 from Kistler Instrumente AG,
which is originally designed to measure cutting forces in
milling. It measures up to 3000N with a resolution of +/-
1N. The sample rate for all tests was 10.000Hz. The
coordinate system is defined according to x and y as the
plane and z orthogonal to the plane, see Figure 6.
around zero. These curves correspond to radial cutting
forces caused by the orbital movement in the orbital
mechanism, see Figure 7. The radial forces were
constant at around +/- 5N independent of both the feed-
rates and hole diameters tested. The axial cutting force
measured forces between 30-40N. Interestingly the axial
force did not rise particularly due to either increased
feed-rate or larger hole diameter.
Conventional Drilling

Z 1
2
X
Y conv_05

Section 1 enters
material
Figure 6: The coordinate system of the Kistler force
sensor Section 1 fully
exits material
The tool changing between the drilling machine and
other equipment used the Capto system from SANDVIK. Section 2 enters Section 1 start
Capto is generally used for holding cutting tools in NC- material exiting material
machines, but worked perfectly as a robot tool changer.
This simplified changing between the conventional- and
orbital drilling machines both in the NC-machine and the
Figure 8: A representative result from conventional
robot.
drilling with a hole diameter: 6 and feed rate:
0.7mm/s
For the first experiments for gathering reference
measurements the orbital and the conventional drilling
The conventional drilling process (Figure 8) had lower
machines were rigidly attached to an NC-chuck. The
radial forces but much higher axial forces. The axial
NC-machine was used to change position of both drilling
forces were between 120 to 320N depending on feed-
machines and specimens during the tests. Several tests
rates or hole diameters. A more thorough investigation
for each method were undertaken, so that a repetitive
of the process in conventional drilling is presented by
outcome of the result was achieved.
Nichol [12].
Orbital Drilling

Cutting edges
on both side
and face

Figure 9: Comparison between different feed rates
for both methods
There was a clear trend that the forces in orbital drilling
did not change when the feet rate was increased, see
Figure 9. The experiments showed that the orbital
drilling can be as much as 8 times lower than
conventional drilling. This trend becomes more evident
when the feed-rate or hole diameter is increased.
DISLOCATION MEASUREMENTS IN THE ROBOT
The same force sensor as in the reference
measurements was used with the robot, which was
attached to a stiff frame. The robot used for the tests
was an IRB4400 from ABB with S4Cplus controller. In
the first attempt the orbital drilling unit was held freely
in the robot TCP (see figure 10). This means that the P
ODU was neither attached to the workpiece or a drilling
template.
Figure 10: The P ODU attached to the robot
The force measurements showed similar behavior as in
the reference. Forces in the axial direction were only
slightly lower when drilling with the robot compared with
the reference measurement.
Dislocation was measured with three LVDT sensors
(Linear Variable Differential Transformer) as shown in
figure 11. They had a resolution of 0.001mm and a
range of +/-2,5 mm. The deflection was measured in X,
Y and Z, which corresponds to the axial, horizontal and
vertical direction. The L-profile was applied to the front
end of the end-effector.
X
Y
Z
Figure 11: The LVDT-rig. X is the axial direction, Y is
the horizontal direction and Z is the vertical direction
The result from the deflection measurements is shown in
figure 12. The axial movement started out with an offset
of 0.06mm followed by a sinusoidal movement with an
oscillation of +/-0.005mm. This can be derived from the
fact that the forces detected from the reference
measurement clearly stated that the axial forces were
greater than the radial forces. However, having the end-
effector held freely resulted in a rather large horizontal
and vertical movement compared to the relationship
between the forces in the reference measurements.
mm
X
Orb_d_01
Sample (10.000Hz)
Figure 12: Typical appearance of deflection in the
robot when the orbital drilling unit is held freely in
the robot TCP
The vertical movement started with an offset of 0.03mm,
followed by an oscillation of +/-0.01mm. This can be
explained from the fifth axis of the robot that deflected
more than the others. This is a result of having the end-
effector held horizontally resulting in a distance between
the axial force and the fifth axis rotation point. The
deflection in the horizontal direction gave +/-0.02mm
around zero. This is more logical when there is
symmetry on both sides of the end-effector. This can be
seen in figure 10, where the specimen is positioned in
the robots horizontal center.
THE LRPVS-CONCEPT
When the robot held the orbital drilling machine without
any attachment to a jig or workpiece deflection occurred
during the drilling process, which affected the quality of
the holes. In order to zero out any horizontal and vertical
movement an additional feature to the drill end-effector
was developed, called Linkping Rubber Press and
Vacuum Suction (LRPVS). The idea was to install a
rubber ring in the end of the end-effector. The concept is
shown in figure 13. By slightly pushing with the robot
against the workpiece the friction stiffened up the drilling
machine against the workpiece.
Vacuum boost
attachment
Rubber
ring Vacuum
Figure 13: A picture of the LRPVS concept
As shown in figure 14, the forces during the drilling cycle
looked a bit different from previous drilling trials. It was
possible to detect the pre-force caused by the robot
pushing against the workpiece. Interestingly 50% of the
axial cutting force was taken up by the robot.
pre-force
Figure 14: Force measured with the LRPVS concept
with 0.5mm/s, 6 mm hole diameter and a 4 mm
cutting tool
Figure 15 shows results from the deflection
measurement using the LRPVS-concept. The movement
in Z (vertical direction) became close to zero. This
means that the LRPVS reduces some of the deflections.
However in the Y-direction (horizontal) there is still some
movement. This can be explained in that the rubber ring
is too thick. In order to reduce the thickness of the
rubber a flange in the orbital drill unit had to be removed.
In an optimal application the rubber thickness should be
as thin as possible to reduce the inherent movement of
the rubber.
mm
X
Z
Y
Orb_d_05
Sample (10.000Hz)
Figure 15: Deflection measurements in the robot
using the LRPVS-concept
It is also important to understand that the LRPVS
solution requires more testing and trimming before
industrial implementation. The movement of 10m in Y
can be zeroed out if the rubber would be thinner and if
the pre-force (robot pressing against the plate) is
optimized. The reason for the movement in the x-
direction (axial direction) comes from not having trimmed
the pre-force. The pre-force must be at least equal to the
axial force arising in the drilling process, which was not
the case in the tests above. This requires further
experiments.
CONCLUSION
This paper has focused on evaluating the cutting forces
in orbital drilling when compared to conventional drilling
and how this affected the result when performing the
same experiments with an industrial robot. The results
show that cutting forces in orbital drilling were as much
as eight times lower when compared to conventional
drilling. Cutting forces when drilling with standard
industrial robots have always been considered a
problem due to robot deflection. Therefore orbital drilling
could be considered as an enabling technology for low-
cost automation of robotic drilling. This paper focused
primarily on orbital drilling with robots and a separate
paper concentrating on the cutting forces in orbital
drilling will be presented in the future. Another feature
derived from the force measurements was that the
variation in the forces when the orbital tool first exits the
material could be used for detecting a new subsequent
layer in a material stack.
ACKNOWLEDGMENTS Conference on Advanced Robotics and Intelligent
Automation, Athens, Sept., 1995
This work is part of the EU-founded project ADFAST 9. Gooch, R., Optical metrology in manufacturing
(Automation for Drilling, Fastening, Assembly, Systems automation, Sensor Review 1998 vol. 18 nr. 2
Integration, and Tooling). Special thanks go to project 10. Horst A. Beyer, 3D Image Metrology for Lean
manager John Andersson. Manufacturing, SAE Aerospace Automated
Fastening Conference & Exposition, 1999
REFERENCES 11. R. Lindqvist, I. Eriksson, M. Wolf, Orbital Drilling of
Sandwich Constructions for Space Applications,
1. George N Bullen, The mechanization/automation of SAE Aerospace Automated Fastening Conference &
major aircraft assembly tools, Production and Exposition, 2001
Inventory Management Journal; Falls Church 12. Nichol, W., Understanding Exit Burr in the Drilling
Quarter 1997 Process, SAE Aerospace Automated Fastening
2. Dgoulange, P. Dachez, F. Pierrot, P. Prat, Conference & Exposition, 2001
Determination of a reference model for controlling
the deformation of an industrial robot, Application to CONTACT
riveting in Aeronautics 1994
3. Whinnem, E., Nystrom, M., Integrated Metrology & Henrik Kihlman
Robotics Systems for Agile Automation, SAE Production System
Aerospace Automated Fastening Conference & Linkping University
Exposition, 2000 Phone (+46) 13-288 974
4. Horng, S., Optimization of Robotic Drilling System, E-mail: Henrik.Kihlman@ikp.liu.se
Robotics & Conference. Detroit, Michigan, 1984
5. Hartmann, J., Automated wing panel assembly for Ingvar Eriksson
the A340-600, SAE Aerospace Automated Novator AB
Fastening Conference & Exposition, 2000 Phone (+46) 8-673 5439
6. Fernandes, M., Gower, S., De Boer, F., Cook, C., E-mail: ingvar@novator.nu
Robotic Drilling of Carbon Fibre, Technology
Convergence in Composites Applications, UNSW, Mark Ennis
Novator AB
Sydney, Australia, 6-9 Feb., 2001
Phone (+46) 8-673 5451
7. Ribere, B., Guihard, B., Bureller, M., High Accuracy
E-mail: mark@novator.nu
And Reaming Machines Used On A A 340-500/600
Final Assembly, SAE Aerospace Automated
ADFAST
Fastening Conference & Exposition, 2001 http://www.euadfast.com/
8. Nikoleris, G., Robot calibration methods in industrial
applications, Proceedings of ECPD International
 Appended Publication V

Kihlman, H. and Loser, R., (2003) "6DOF Metrology-integrated Robot Control", SAE
2002 Transactions Journal of Aerospace, ISBN Number: 0-7680-1448-4, pg 398

6DOF Metrology-integrated Robot Control
Department of Mechanical Engineering, Linkping University
Copyright 2003 SAE International
ABSTRACT
This paper describes ongoing research into Metrology-
integrated robot control. The research is a part of an
ongoing EU funded aircraft industry project ADFAST*.
The ADFAST project tries to implement the use of
industrial robots in low-volume production, high-demand-
on-accuracy operations and for dynamic force
compensation. To detect and compensate deflection in
industrial robots during a process, the robot uses a
metrology system. The metrology system supervises the
tool center point of the robot as it executes its
processes. Leica has recently released a new metrology
system; the LTD800, which measures distances with
laser interferometry and can simultaneously measure
orientation of targets, through photogrammetry, using an
additional camera on top of the measuring unit. This
paper will describe theory and results from tests
performed on integrating the LTD800 with the robot.
INTRODUCTION
Using robots in aircraft manufacturing has been common
practice for many years. Industrial robots, sometimes
called lightweight robots, have two main problem areas:
low absolute accuracy and stiffness. The aerospace
industry, if compared to automotive industry, has low-
volume production, typified by a large number of
operations to perform but little repetition. One example
of this is the drilling of holes in an aircraft structure prior
to fastening in the assembly process. Using robots for
this purpose requires offline-programming and depends
on the absolute accuracy of the robots to stay within the
tolerance span of desired product quality. Even though a
well calibrated industrial robot today could achieve the
absolute accuracy needed, there is still a big risk that it
will be dislocated or deflected if the forces in the process
performed are changed. This is indeed the case for a
drilling process, which takes up most of the areas for
automation in the aerospace industry. If, however, an
external measuring system were to be used to track
the position of the TCP continuously during the drilling
2003-01-2961
Henrik Kihlman
Raimund Loser
Leica Geosystems AG
process, the resulting deflection could be identified and
corrected.
In a part of an ongoing EU-funded aerospace industry
project - ADFAST*, a newly developed metrology
system, the LTD800 from Leica, has been integrated into
a conventional industrial robot, the IRB4400 from ABB
Inc. The controller of the robot is operated through a
system called WebWare, which enables the robot to be
controlled from an external PC. The servo controller in
the PC interacts with the LTD800 through an embedded
system called emScon. A demonstrator, called DEMO-
METROLOGY, has been developed in ADFAST to
evaluate the ability to integrate the LTD800 with the
robot in real-time. One of the objectives has been to
compensate for inaccuracies in the robot during an
aircraft assembly process. The theory and result from
the DEMO-METROLOGY is presented in this paper.
ROBOTS IN AIRCRAFT ASSEMBLY
Using robots to automate aircraft assembly is not new in
a general sense. In the ADFAST project, the objective is
to use an industrial robot for automation of scenarios like
low-volume production, high-demand -on-accuracy
operations and dynamic force compensation. The
following section will describe the different focus areas of
this project, and how the robot has been utilized.
ROBOTIC DRILLING
The largest application area for industrial robots is still
the automotive industry, where robots are primarily used
for performing repetitive operations. In this case, one
robot program is generated and often used for a number
of years. In the execution of these programs, the robot is
calibrated for a required level of performance within the
repetitive pattern it moves. If new operations are to be
performed, more calibration is required to ensure
position accuracy for the new pattern. Recent attempts
to reach more exact positional accuracy have been
undertaken (Helin et al. 2002). One of the obstacles with
absolute accuracy is that it can only be achieved in a
static load configuration. If however, the forces applied THE LTD800 METROLOGY SYSTEM
on the robot TCP changes, the absolute accuracy in the
calibrated robot can no longer be insured. This is always The LTD800 system, together with its 6DOF
the case when a robot is used for automating drilling. accessories, is used to provide the measurements to
DeVlieg et al. (2002) presented a number of problems guide the robot at the final position. The system includes
encountered by robots during drilling, to include a Laser Tracker with a control unit, a high-speed camera
imperfect kinematic models, payload drop, backlash, and a specific target device, and is essentially a
flawed TCP definition and imperfect gear rations. combination of tracker technology and photogrammetry.
Putting those two metrology tools together will not only
Drilling has two courses of events: the first one is to allow for measuring a point in space in real-time, but
apply the drilling machine to the position where the hole also for orientation in space as shown in figure 1.
is to be drilled; the second one is to hold the drilling
machine in that position throughout the drilling cycle. In
order to hold a drilling machine steady during drilling, a
certain pressure must be applied by the robot between
the structure to be drilled and the drilling machine prior
to drilling (Kihlman et. al, 2002). This pre-pressure
causes the robot to deflect, since the robot can be seen
as a spring model (Loncaric, 1985). Riving (1988) stated
that 70% or more of the compliance in a robot is in its
joints. The context from this is that something else is
required to help the robot to compensate for the
deflection of the drilling end-effector after pre-pressure
has been applied and to supervise the position of the
drilling machine while the hole is drilled.

ROBOT MANIPULATED FIXTURES

Figure 1: Laser Tracker combined with
Limited research has been undertaken in the use of the Videogrammetry.
robot to perform the reconfiguration operation of flexible
tooling (Youcef-Toumi and Buitrago 1988; Goodman, MEASURING PROBES
1990; Kihlman, 2002). In this research, the robot docks
onto fixture parts and changes the configuration of the Measuring orientation is achieved through the use of a
tool in order to handle different product types. This high-speed digital camera to simultaneously image all
approach involves many challenges. The forces exerted diodes on a probe. The probe shown in figure 2 is called
on the TCP, for example, are seldom gravitational, and a T-probe, which is a hand-held probe that is suggested
are therefore forces that the robot in its controller cannot to fit into a spatial envelope of typical dimension 20cm.
compensate for, which lead to uncontrolled deflection. In
addition, when the fixturing modules are reconfigured,
friction occurs. This friction will vary depending on
configuration and speed. Positioning fixture modules are
coherent to the quality of the product; hence, the
demand on positional accuracy of fixture parts is high.

When considering robot automation in the aerospace

industry today, these problems are not always the case.
If a robot of sufficient size is used, and if the strength
and stiffness of the manipulator is large enough, the
friction of the modules being manipulated can be largely
negated. One fully working implementation of robot-
manipulated tooling with a larger robot unit was
presented in a project by Munk (2002). However, if small
and cheap off-the-shelf industrial robots are to be used
with payloads less than 100kg, the friction and weight Figure 2: The T-probe for manual measurement
cannot be neglected. To supervise unexpected
deflections in these kinds of robots and to reach the The hand probe can be used for single offset probing of
product quality requirements, an additional external points. It can, like a CMM touch probe, reach into objects
measurement system is necessary to ensure the in order to locate features not directly visible to the laser
positional accuracy of the robot. tracker. The 6DOF probe concept is thoroughly
explained in Loser and Kyle (2003). The T-probe is, in
this paper, only used for explaining the functionality of
the LTD800. In the experiments performed with the
robot, however, the 6DOF Metrology Probe is used
(explained in its own section later).

The positioning of the reflector is determined with the

Laser Tracker, which operates within 10 ppm (m/m).
The orientation angles (pitch, roll and yaw) can be
determined with 0.01-0.02.

By using a glass prism retro-reflector, the tracking laser

beam can be some 50 away from the prism axis before
contact is lost. This provides an operator with over 100
of flex ibility in movement. This effect comes from the
refracting effect of the glass, which ensures the effective
target point is no longer the fixed intersection point of the
reflecting surfaces but a shifted virtual point whose shift
is dependent on the angle of the laser beam to the
prism. The maximum extent of the LEDs in the probe is
20cm with a depth distribution of 95 mm, which is
appropriate for the angular sensitivity required.

A distinguishing feature with the probe is that the tracker

can send comman ds to the probe by modulating the
signal of the Absolute Distance Meter (ADM) (Loser,
1996). The ADM beam is co-axial with the interferometer
beam and is used to re-initialize the interferometer range
after the beam is lost or interrupted. The tip of the prism
apex in the probe has been ground down to produce a
small hole, through which part of the ADM beam passes
through and is detected through a light-sensing device
whose output corresponds to the beam modulation. This
means that the probe, if battery-pow ered, can be
completely wire-less.
LASER TRACKER LTD800
The Laser Tracker represents the central measurement
device, and is able to track a reflector and measure its 3-
D position in space with an update rate of 1 kHz. The
internally used interferometer performs precision
encoding, and the position detector guarantees a high
measurement accuracy of 10 ppm. To avoid long re-
initialization procedures if the interferometer beam is
interrupted, an absolute distance measurement device is
integrated to reset the interferometer distance. In
addition to this basic measurement capability, a high-
speed camera, the T-Cam, can be mounted on top of the
Laser Tracker to measure the 3 orientation angles, see
figure 3.
Figure 3: The T-cam
CAMERA SYSTEM T-Cam
The T-Cam uses a vario lens for a fixed lens in the
camera. Sensitivity to change in probe orientation is
better if the targets fill more of the image space. As the
range of the probe increases, the usable image space in
a fixed focus lens will decrease, leading to reduced
accuracy; hence, the camera is equipped with a tracking
vario lens which changes the focus and maintains
constant scale through an operating range of 1.5m to
15m.
CONTROL UNIT LTC-plus
The LTC-plus represents the brain of the systems, and
coordinates and controls all systems activities.
Therefore, a triggering and synchronization board
provides the time representation, and measuring events,
marker flash time and the electronically shutter of the
camera.
Communication with the LTC-plus is done trough TCP/IP
with the standard command: SendData(). When the
ReadData() is triggered, data is red back to the
application.

THE EMBEDDED SYSTEMS

In the implementation of the 6DOF Metrology-integrated

Robot Control, two embedded systems were used; the
emScon was used to integrate the LTD800 to the servo
controller, while the WebWare used for integrating the
robot control to the servo controller on the external PC.
This section will briefly present these two systems.
EMSCON

Through the Embedded System Control (emScon), the

Leica Laser Tracker is completely integrated into the
manufacturing process, giving the ability to integrate the
Tracker into machining guidance and the calibration
process. With the new Leica Tracker Programming
Interface, it is possible to create modules for any
application software to fully control the Leica Laser
Tracker out of world standards like Excel, Word, Access,
Visual Basic or C++. Since there is no dependency on
operating platforms, it is possible to use Windows, Unix,
or Linux, among others to control the LTD800.

The Tracker Programming Interface can, in addition to

the LTD800, also be used with the rest of the Leica
Laser Tracker Family, such as the LT300, LT500 and
LTD500 and other future versions.

WEBWARE

In the application presented in this paper, an interface

called WebWare SDK has been used for communicating
Figure 4: The topology of the system
between the ABB robot controller and the external PC.
The WebWare SDK supports 32-bit Windows
The PC ran a Windows 2000 operative system with two
applications created with either Microsoft Visual Basic,
additional embedded systems: emScon and WebWare,
Visual C++, or scripting. The WebWare SDK interface
which where implemented using Visual Basic.
consists of a set of OLE Custom Controls known as
Communication was set-up by a TCP/IP communication
ActiveX controls. The ActiveX interface mechanism is a
protocol through an Ethernet switch.
Microsoft standard utilized in all Microsoft visual
programming environments. The WebWare SDK
presents the entire robot controller communication In the PC, a servo controller was utilized with a generic
architecture to adapt to new operations from the
interfaces in a manner consistent with other common
tools developers use to create Windows applications. operation planning system. In the operation planning
system, new instructions are generated depending on
the task to be performed. The servo loop executes
SYSTEM TOPOLOGY different sub-routines depending on what is to be
executed. A module-based architecture enabled new
The DEMO -METROLOGY developed in the ADFAST sub-routines to be created to adopt for changing
project was built in the robot laboratory at Linkping conditions.
University. The DEMO-METROLOGY consisted of one
conventional industrial robot, one laser tracker and an
external PC. The robot was an IRB4400 from ABB with a
S4C+ controller. The laser tracker was an LTD800, with
the CONTROLLER plus version with an integrated
emScon server package. Figure 4 shows an overview of
the topology of the system.
 Metrology Probe to the robot. Figure 6 shows the 6DOF
Metrology Probe.
Servo
LTD800 Controller IRB4400
Base pos. GripLoad
Unit spec Base pos.
TCP pos.

Sub
Drilling routines Tooling
Machines Adjust Modules
Reconfig
#1. #1.
- Cutter Rebuild - Base pos
- Feed rate Dock - Top pos
- -
#2. Static calib #2.
- Drill -
- Error handl -

Figure 6: The Metrology Probe for the robot

Figure 5: The architecture of the server controller on
the external PC The 6DOF Metrology Probe can be used for three
different purposes. The primary use is to be attached
Through the use of embedded systems, such as beside the robot chuck that is used for attaching drilling
WebWare and emScon, all code running in the system is machines and for interfacing tooling modules, see figure
done only in the servo loop. When the Tracker is called 7.
for measurements, an embedded object is initiated. The
embedded object performs a conventional socket
communication execution to the emScon server, and a
response is given. This was also done when executing Chuck
the robot program. If the robot is instructed to drill a hole
on a certain point, a sub-routing called Drill is called, and
the robot instruction MoveL in the RAPID language is
called directly from that sub-routine. Hence, there is no
dedicated program needed for the robot to perform such
an operation, but instead a generic program, executed
from the servo controller through WebWare, which is
referred to as having the robot execution embedded in
the servo controller in the external PC. Figure 5 shows
an overview of the servo controller.

THE 6DOF METROLOGY PROBE

The 6DOF Metrology Probe is the unit that is attached

close to the TCP of the robot, referred to from this point
as The Metrology Probe.
The basic functionality of the Metrology Probe is to carry
the reflector and the marker LEDs in a fix and stable
relation to the mounting device. As first test, a standard
Capto chuck from SANDVIK inc. with one male and one
female connection, normally used as a fixture device for
attaching cutting tools in NC-machines, was integrated
into the Metrology Probe. The Capto reference system
from allows fixation with a repetitive accuracy of 2m
and can carry high loads. According to the T-Probe
described above, the markers LEDs have a similar 3-D
distribution. The Metrology Probe consists basically of
the same electrical and optical components as the T-
probe, but the housing is different with the Capto
interface attached in the center of the probe to attach the
Figure 7: The Probe attached beside the Robot
Chuck
The second purpose is to calibrate the chuck position
relative to the robot TCP, whereas the Probe uses the
male interface to be attached to the robot chuck that has
a female interface. Before attaching the probe beside the
chuck, the chuck TCP is measured. When the Probe is
attached besides the chuck, the relative position
between the chuck TCP and the position beside the
chuck can be measured, and later used in process
operations through a simple transformation matrix.

Figure 8. The Probe attached to the Robot Chuck
The third purpose for the Probe is to calibrate the
position of all moveable modules in the flexible tooling
system developed in the ADFAST project. In the end of
each tooling module a male Capto interface is attached,
which has three purposes. The first purpose is to allow
the robot to dock onto the module and reconfigure it.
Secondly, as an attachment for product specific pick-ups
in the assembly tool. Thirdly, for allowing the Metrology
Probe to be attached, see figure 9.
Figure 9: The Probe attached on a Dynamic Tooling
Module of a flexible tooling system
When all Dynamic Tooling Modules has been calibrated,
the robot knows their exact position and can dock onto
them and reconfigure them for a new change-over. More
about this tooling system can be red in (Kihlman, 2002;
Kihlman and Engstrm, 2002).
EXPERIMENTAL RESULTS
This section presents the results undertaken at current
stage in this research.
POINT-TO-POINT CONTROL
This experiment was aimed at testing the system to go
from one point to another, and no tracking was done
between the points, however when reaching the point,
the robot informs the servo controller about its position.
The tracker points the laser beam to that position and
identifies the reflector. When the tracker finds the
reflector, the ADM is used to measure the position of the
reflector. When the ADM has calculated the distance to
the reflector, the T-Cam adjusts the camera lens to
zoom in the leds, whereupon the orientation can be
calculated. At this stage the tracker is now locked to the
Metrology Probe. From this point the Servo Controller on
the external PC iterates the robot to until the correct
position is reached. The iteration stops when the robot is
located within tolerance. The test runs proved the robot
to be located within 0.1 mm.
This test showed an important phenomenon. A normal
industrial robot is programmed point-to-point. This
means, when the robot is given a position to go to (i.e. a
robot target), the position cannot be manipulated from its
main program, since the robot locks the program until
the position is reached; see Alternative 1 in figure 10
below. The solution to that problem is solved by using
the ability to use multitasking in industrial robots. By
updating the robot target using a multitask object in the
robot, the robot can update its target before reaching it,
see Alternative 2 in figure 10.
Alt 1. First point
x x New point
x Alt 2.
x start The robot is given a new
position
Figure 10: Two importance with using multitasking
in the robot
It is important to understand the difference with being
able to controlling the robot in real-time or not. If
multitasking where not used, the regulator would be
much slower and give strange behaviors. The multitask
object in the robot is used to communicate with the
Servo Controller on the external PC, so that the
regulator on the Servo Controller can update the robot
target during movement.
ONGOING EXPERIMENTS
When this paper was written, all test result had not yet
been compiled due to delays and technical problems in
some of the equipment. Final results from the suggested
experiment presented in this section are to be presented
at the SAE conference, where this paper was published.
Besides the already presented Point-to-point control,
four other experiments are to be fully implemented in the
DEMO-METROLOGY installation. The tests are
presented in this section.
CONTINUOUS CONTROL
Continuous control means the LTD unit is continuously
measuring the position and orientation of The Metrology
Probe when the robot is moving. This is sometimes
called tracking, and compared to the point -to-point test;
the robot position is supervised by the tracker and
adjusted continuously, not just to iterate the robot to one
point. This test is the most challenging one, because the
speed that the robot moves in is dependent on both how
quickly the LTD can measure data and how quickly that
data can be used for compensation in the robot. This is
called synchronization. This test indeed identifies the
bottlenecks in the system.
As mentioned earlier, the robot is not always aware of its
deviation from the nominal point. Using an external
metrology system, this deviation can now be detected.
Putting it in a philosophical perspective, it is like giving
the robot eyes to see what is happening. The servo loop
on the external PC can now update the robot target
continuously to make the robot reach closer to the
nominal location.
FIXTURE RECONFIGURATION
As mentioned earlier, the robot in this research is used
for reconfiguring a flexible fixture. In this scenario the
load can change randomly, which makes it impossible to
instruct the robot controller with what the current forces
are. Reconfiguring these units also involves friction,
which changes depending on speed and orientation of
the units. All together, these experiments are the first
steps towards handling dynamic loads that are
unambiguous and that would be impossible to
compensate for using a lightweight robot without an
external regulator.
DRILL UNIT SET-UP
This part was mentioned earlier. Basically, this
experiment was about answering the questions of how
could a lightweight robot compensate for the compliance
in its structure when applying a pre -pressure prior to
drilling. The experiment is rather straightforward. The
robot is instructed to position the end-effector against a
surface.
Presented by (Kihlman et. al, 2002), deflection will occur
during this operation, where the TCP is dislocated from
its nominal position. In steps the robot is forced to
compensate the misalignment of the TCP. This is
executed in a separate sub-routine in the Servo
Controller presented earlier in this paper. The modularity
of the system simplifies this experiment, since it is easy
to go in and work on this particular part of the program.
DRILLING SUPERVISION
This sub-routine is basically a supervising operation. If
the pre-pressure is correctly arranged, the robot shall not
move during drilling. However, the supervising process
is recording the drilling process and used for analyzing
hole quality. If there is movement during drilling, the
quality can be compromised, and the operator is alerted
to verify the quality of that hole after it had been drilled. If
the deviation were too high, the drilling cycle would go
into reset and call the Error-handler. In this case, the
question is arisen: is it possible to avoid pre-pressure
prior to drilling, and instead compensate for the dynamic
forces during the drilling cycle? The answer to that
question at this stage in the research is no! The changes
in forces are much too fast than what could be
compensated for in real-time by the Servo Controller. In
that case a quicker connection to the robot controller
would be required. Today, the TCP/IP communication
sets the limit of which speed that can be reached.
CONCLUSION
Although not all results from the experiments presented
in this paper could be finalized in time, the integration
between a lightweight industrial robot and a metrology
system has shown promising results. The multipurpose
Metrology Probe has proved to be a cheap and trustful
device to quickly calibrate the 6D-position of a flexible
tooling system, to supervise robot drilling and to improve
positional accuracy of a robot during process, hence,
reducing the conventional strenuous calibration work
that seldom allow for shortcuts.
ACKNOWLEDGMENTS
This work is part of the EU-founded project ADFAST
(Automation for Drilling, Fastening, Assembly Systems,
Integration and Tooling). Special thanks go to the project
manager of the ADFAST project, John Andersson. We
would also want to thank SANDVIK inc. for showing
great interest in using the Capto system in this research.
REFERENCES
1. DeVlieg, R., Sitton, K., Feikert, E. and Inman, J.
(2002) ONCE (ONe.sided Cell End effector) Robotic
Drilling System. SAE Aerospace Automated
Fastening Conference & Exposition, October 1-3
2. Eugene, I., Rivin, (1988) Mechanical Design of
Robots. ISBN: 0070529922
3. Goodman, H., (1990), Assembly cell speed riveting,
Tooling & Production, December
4. Helin, P., Jerregrd, H., Robertson, A., and Snell, J. 11. Munk, C. (2002) Determinant Spar Assembly Cell.
(2002) Technologies that make a robot reach SAE Aerospace Automated Fastening Conference &
Absolut Accuracy, Proceedings of the 33rd Exposition, October 1-3
International Symposium on Robotics, October 7-11 12. Youcef-Toumi, K. and Buitrago, J. H. (1988) Design
5. Kihlman, H. and Engstrm, M., (2002) Affordable of robot-operated adaptable fixtures, Proceedings,
Reconfigurable Tooling, SAE Aerospace Automated Symposium on Manufacturing Systems - Design,
Fastening Conference & Exposition, October 1-3 Integration and Control, Manufacturing International
6. Kihlman, H., Eriksson, I. and Ennis, M. (2002) Conference, Atlanta, Ga., pp. 113-119
Robotic Orbital Drilling of Structures for Aerospace
Applications. SAE Aerospace Automated Fastening CONTACT
Conference & Exposition, October. 1-3
7. Kihlman H. (2002) Affordable Reconfigurable Henrik Kihlman
Assembly Tooling - An Aircraft Development and Department of Mechanical Engineering
Manufacturing Perspective, Licentiate Thesis No. Linkping University
980, LiU-TEK-LIC-2002:53, Department of SWEDEN
Mechanical Engineering, Linkpings Universitet, SE- E-mail: Henrik.Kihlman@ikp.liu.se
581 83 Linkping, Sweden, ISBN 91-7373-460-8
8. Loncaric, J. (1985) Geometrical Analysis of Raimund Loser
Compliant Mechanisms in Robotics. PhD in applied Business AREA IMS
mathematics. Cambridge, Massachusetts Leica Geosystems AG
9. Loser, R. (1996) The Authority of Common Unterentfelden
Measuring Task Through the Integration of ADM SWITZERLAND
with Laser Tracker, Coordinate Measurement E-mail: raimund.loser@leica-geosystems.com
Systems Committee, Williamsburg Hospitality
House, Williamsburg, Virginia, July 8 - 12 ADFAST
http://www.euadfast.com
10. Loser, R. and Kyle, S. (2003) Concepts and
components of a novel 6DOF tracking system for 3D
metrology, Optical 3D Measurement Techniques
Conference. Zurich, 22 -25 September
 Appended Publication VI

Kihlman, H., Sunnanbo, A., Loser, R., Von Arb, K., Cooke, A., (2004) "Metrology-
integrated Industrial Robots Calibration, Implementation and Testing", 35th
International Symposium on Robotics, Paris-Nord Villepinte, France, March 23-26
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004

Metrology-integrated Industrial Robots

Calibration, Implementation and Testing

Henrik Kihlman, Raimund Loser, Andrew Cooke,

henrik.kihlman@ikp.liu.se raimund.loser@leica-geosystems.com andrew.r.cooke@baesystems.com
Albin Sunnanbo, Konrad von Arb,
Albin.sunnanbo@ikp.liu.se konrad.von-arb@leica-geosystems.com
Linkping University Leica Geosystems AG BAE SYSTEMS
Dept. of Mechanical Engineering Metrology Division Advanced Technology Centre
SWEDEN SWITZERLAND UNITED KINGDOM
ABSTRACT model. Attempts to use robots for more demanding
This paper presents integration of a metrology system operations, such as drilling has been undertaken in
and an industrial robot. The metrology system this research with success. Another demanding
consists of a laser tracker that measures the distance operation studied in this research and presented in
to a prism with high accuracy and a camera that this paper is to use the robot to perform
through photogrammetry measures the orientation of reconfiguration in a flexible fixture. In order to reach
a reflector. Both laser prism and camera reflector is and maintain high accuracy, this paper is presenting
integrated to a 6D-Reflector that is attached close to an approach to have a metrology system online with
the TCP of an industrial robot. Tracker and robot is the robot controller to compensate for errors
connected to a PC on a TCP/IP network. The PC continuously. This removed much of the calibration
takes measurements with the tracker, and thereby work, since most of the calibrations were performed
compensates the robot to reach high absolute online. The metrology system uses a laser tracker to
accuracy in the robot positioning (+/-50 m). The measure the position of the robot Tool-Center-Point
6D-Reflector has multi-functionality and simplifies (TCP) and an additional camera to measure
calibration procedures. This paper explains the orientation. Tests have been undertaken where the
architecture of the system and the methods for robot reaches 0.05 mm positional accuracy and 0.04
calibration. degrees in angular accuracy. This paper includes an
introduction to the hardware used in the installation,
Keywords: metrology, laser, photogrammetry, robot, calibration methods and results from the integration
online control, calibration tests. This paper is a continued session from an earlier
published by Kihlman & Loser, 2003 (6). In the
1 INTRODUCTION previous paper, due to technical problems and delays,
all the final results could not be published in time.
Industrial Robots have been developed from the This paper covers these results plus more.
beginning essentially for automating welding in the
Automotive Industry and has been used for many
years with success. In the Aerospace Industry 2 ROBOT ACCURACY
automation is much more expensive for several Industrial robots have for many years been used in
reasons. Two major reasons are the low product great extent in Automotive Industry. Aerospace
volumes, and high demands on accuracy. Automation Industry is now prudent to reduce cost and shorten
in Aerospace Industry has mainly been accomplished lead-times, where large dedicated machines have
through highly dedicated machines. The level of been the common method for automation. These
interest in using Industrial Robots for automation in machines have been a guarantee for enabling high
Aerospace Manufacturing has increased in recent accuracy in difficult operations, such as drilling and
years. On the market today, there exists several assembly of high quality products. These machines
examples of High-Accuracy Robots. High-Accuracy however, are expensive and lack flexibility.
Robots are subject to a calibration phase prior to Therefore, Aerospace Industry has now realised the
delivery. This provides an absolute accuracy in the potential of using industrial robots for automation.
order of 0.5 mm or in some cases even higher. This Automotive Industry has also put out new demands
method requires the robot process load to be in on the robots to handle new tasks, which requires
direction of gravity and static, so that the robot higher precision. Amongst others, these two scenarios
controller can maintain the accuracy. On the market have led to robot suppliers increasing the capability
there exist methods for cell-calibration, where robot on their robots; hence today robot providers are
and end-effectors are calibrated to perform certain presenting new robots with high accuracy programs.
tasks that requires higher accuracy. These methods
however, require the process forces to remain the 2.1 Resolution in robots
same over time. In some processes the conditions are
difficult to predict, such as processes with dynamic What makes robots reach high accuracy using a
loads i.e. drilling, or friction when manipulating metrology system is neither repetitive- nor absolute
flexible fixtures. These scenarios could be solved if accuracy itself. What is important is the resolution of
there existed perfect models of the system. These the robot. Resolution on the other hand often goes
systems however are complex and are difficult to hand in hand with high repetitive accuracy. Let us
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004
look at the robot used for the experiments - an
IRB4400 from ABB. The repetitive accuracy in that
robot is 50m. To reach that accuracy in the full work
volume, the resolution of the robot is much higher.
Tests performed in this research have proven the
robot to able to move in a resolution of 5um. When
fine adjustments have been carried out, the robot
controller is simply given a new absolute position in
the base coordinate system. Helin, 2002 (3) presented
tests results from ABBs High-accuracy program.
The results clearly showed that it is easier to reach
higher accuracy in smaller robots. One trend in
Aerospace Industry nowadays is to buy large robots
for being able to carry multi-end-effectors weighting
several 100kg. It is contradictory to use large robots
when high accuracy is required. This indicates that
end-effector suppliers should consider building
smaller end-effectors if they are to be fine adjusted,
which requires smaller robots and the question arising
to the end-effector suppliers today is: How will large
end-effectors be positioned with high accuracy? To
give better understanding to the background to this
research, the next two section will presented why
extensive calibration procedures will not help to
maintain high precision in some scenarios.
2.2 Dynamic loads
Drilling using an Industrial Robot is an example
where
Joint forces
5 change rapidly. Today, suppliers for
drilling end-effectors handle this problem either by
pre-pressurising the drill bushing, or a pressure foot
on the drilling end-effector. This introduces forces to
the robot. To eliminate disturbances on the robot the
pre-pressure force must be significantly higher than
the axial drilling force, to ensure that the drilling end-
effector will maintain its position and orientation
during drilling. One problem here, is that, the
direction of the drilling force, is seldom in the same
direction as a hanging load. Robots today are not able
to compensate for forces other than gravitational.
This will cause problems like slip-stick etc. (1). Also,
when using large and heavy end-effectors it is
common to have the end-effector installed in a
hanging mode on the robot to avoid unnecessary
torque in joint 5 such as figure 1a. When applying
pre-pressure force on the end-effector joint 5 will be
exerted to high forces, in some cases more force than
the weight of the end-effector itself.
Figure 1a Figure 1b
Weakness in joint 5 was presented by Kihlman et. al,
2002 (7). To avoid deflection and slip-stick, the
correct configuration should be to have a pointing
mode, where joint 5 is in zero degrees, see figure 1b.
This gives the robot a chance to increase its base
stiffness to some extent. One has to be cautious for
singularities using the pointing mode. When joint 4
and joint 6 is passing through zero axis degrees, these
kinds of robots will cause strange behaviours.
Pointing mode is also difficult for the robot to
manage if the weight of the end-effector is to high,
which will cause a torque on joint 5. For example, the
IRB4400/60 used in the experiments manages 60 kg
at maximum 27 cm (16.2 Nm) out from the TCP.
2.3 Friction in workobject
Also in this research the robot has been tested as a
manipulator for flexible tooling (4),(5). The robot is
docking with tooling modules and configures them,
see figure 2. The tooling modules presented in that
research can be manipulated in 6DOF. Using the
robot to change the configuration of a mechanism
however is a challenging task, especially in fine
adjustment. There was a clear indication that it is
more difficult to position a robot in high accuracy if it
is exerted for friction. Tests performed in this
research verify this hypothesis. In general, the time to
reach high accuracy becomes longer if the robot is
moving a fixture module. Some tests have also
resulted in overshooting when compensating the
datum point of the fixture module.
Figure 2: Robot-manipulated Tooling
The conclusion from this section is that even highly
calibrated robots would not help in applications
where the process force changes rapidly or if friction,
which is not static, is involved. One solution to cope
with these problems is to have a metrology system
online with the robot. That is what the rest of this
paper will present.
3 EXPERIMENTAL SETUP
Most of the hardware for the metrology integration
has been presented by Kihlman & Loser, 2003 (6),
but to keep this paper together, the components will
briefly be presented in this section.
3.1 The LTD800
Normally Laser Trackers are based on measurements
of a single reflector. Whereas laser interferometers in
general must track the reflector relatively from a fine
calibrated start position (birdbath), if the beam is
broken or another reflector is to measured, the tracker
requires re-calibrating the reflector in the start
position. The LTD800 however, uses an ADM
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004

(Absolute Distance Meter), (8),(9). The ADM enables
the tracker to locate individual target reflectors
without the need to measure from a calibrated origin
position. The ADM can independently measure the
absolute distance to any unknown reflector location,
which simplifies re-initialising the interferometer.
Whereas the ADM is only used in initialising the
interferometer, it cannot itself be used for tracking.
The LTD800 can have an additional camera, the T-
Cam, which is attached on top of the tracker unit. The
T-Cam measures the orientation of a target. The
accuracy of the LTD800 is 10 ppm (m/m) for static
targets, and 20 ppm for moving targets. Distance
resolution is 1.26 m and angle accuracy is 0.02
degrees. The angle accuracy is kept constant over the
full measurement volume, through the use of a zoom
objective in The T-Cam. The zoom objective
continuously maximizes the resolution from the
reflector targets in the camera picture.
3.2 The 6D-Reflector
To measure robot positions in 6 DOF a reflector was
developed. The 6D-Reflector is a further
development of the T-Probe that is a commercial
product from Leica Geosystems AG. The 6D-
Reflector is always initiated with the ADM, since
there is no birdbath available for the 6D-Reflector.
The 6D-Reflector comprises of an aluminium housing
that is attached 10 LEDs and a prism. The LEDs
enables the T-Cam to measure the orientation of the
6D-Reflector.
equipped with an automatic hydro-mechanical Capto
chuck system. This enables the robot to interact with
flexible tooling modules presented by Kihlman and
Engstrm, 2002 (5), ordinary drilling machines and
Orbital Drilling machines such as presented by
Kihlman et al., 2002 (7). The controller on the robot
used for the experiments was the S4Cplus version.
4 COMMUNICATION
In order to connect a metrology system with a robot,
the robot needs to be able to communicate with the
metrology system. ABB robots, which have been
used in this research, have today basically two
communication ports. One is serial port, which has
been a standard for communication with robots in
general for many years. Today exists also TCP/IP,
which now has become a standard for communication
with robots. RS232 and TCP/IP has until now mainly
been used for downloading robot programs from an
offline programming system. Presented in this paper,
TCP/IP has been used for synchronising a servo loop
on an external PC with a metrology system and the
robot controller.
4.1 WebWare
The communication with the ABB robot was
achieved through WebWare. WebWare is a driver
package from ABB that uses ActiveX controls and
OPC (OLE Process Control) server to enable a
network connected computer to manipulate RAPID
program execution, RAPID variable data and I/O-
signals on the S4 controller.

4.2
Male Capto
Female Capto
Figure 3: The Multi-purpose 6D-Reflector
The LEDs are positioned at different depths for the
camera to measure orientation. The LEDs are flashing
in infrared light and the camera is zooming in the 6D-
Reflector so that the LEDs are taking up the full
picture. The laser beam is reflected back to the
tracker unit from a glass prism retro-reflector. The
glass enables the beam to be +/- 50 degrees away
from the prism axis before contact is lost. The 6D-
Reflector housing has two ergonomically designed
handles when moving the 6D-Reflector manually. For
attaching the 6D-Reflector a system called Capto C4
from the Company SANDVIK AB was used. The
Capto system is generally used for attaching cutting
tools in NC-chucks. The repetitive accuracy of the
Capto C4 is 2 m with a maximum load capacity up
to 17 kN.
3.3 The Robot
The robot is a standard Industrial Robot, the IRB4400
from ABB. The IRB4400 has a repetitive accuracy of
50 m and a maximal payload of 60 Kg. The robot is
emScon
For accessing measured data using the TCP/IP
protocol, emScon from Leica was used. emScon is a
tracker-programming interface for complete
integration with Leica Trackers. The emScon server
is accessed over the network through conventional
socket communication. A call for measurement is
initiated, and a response is given. This execution can
either be performed synchronously, where the
program is stopped until a measurement is sent back
to the client, or asynchronously where a trigger is
activated when the data is sent back to the client. The
LTD800, according to specification, is able to
measure 6DOF in 100 Hz and can interpolate
between measurements in 1000 Hz. To use
measurements for feedback control however, requires
each measurement to be sent by emScon to the PC
prior to updating the robot. To clarify this, the
emScon server is sending 4 packages with data per
second. This means that the 100 measurements per
seconds are divided in 4 packages of 25
measurements. For a feedback control loop, only 4
measurements can be used for updating the robot
controller per second, hence the rest is just old data.
This indicates that the bottleneck in the integration
now is the emScon server. Updating frequency in the
system is hence 4 Hz. For the applications presented
in this paper on the other hand, this update frequency
was enough to manage the processes. If however the
robot needs to be updated at higher speeds, such as in
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004
compensating deflection in drilling, this speed would
not be enough.
The LT-controller plus includes two independent
computers. One is emScon that runs on a Windows
2000 operative system, which is not a real-time
system. The tracker processor is running on a real-
time system and could support a high-speed interface.
It could be solved in two ways; either through a
parallel interface using IEEE 1284 that would enable
data transfer with a very short delay and low
overhead. Transferring a measured point would with
this method take only 100-200 microseconds.
Another method would be to use a 100Mbit/s or
1Gbit/s Ethernet with point-to-point connection,
using the UDP protocol. Measurements could then be
transferred within 1-2 milliseconds. Neither of these
methods is supported today, but it could be possible
to accomplish. Changing communication method to
real-time would as a consequence also require
another platform for the Server PC, which today runs
on a Windows 2000 operative system. Transferring
measurements in real-time would be triggered at a fix
rate based on a clock, or a trigger line, whereby
transferring of data is done with only a short delay.
TCP/IP and emScon would still be used on a real-
time platform, but only for things such as setup of the
tracker, starting and initialising tracker and camera,
establish transformation parameter from the tracker to
the object coordinate system etc.
4.3 Integration
The integration was implemented with a standard
TCP/IP network set-up. An Ethernet switch connects
the PC with the robot controller and the emScon
server, see figure 4.
Figure 4: The system architecture for the integration
Since the emScon server is limited to 4 updates per
second, the TCP/IP communication did not limit the
speed. Delays using the TCP/IP protocol did not
affect the synchronisation speed at this time. IP-
networks however, are unreliable due to packet loss
etc., and is not a real-time communication platform.
Tests on round trip time (RTT) were performed. The
test results indicated that RTT was about 30 ms, with
spikes sometimes up to 50 ms. It was clearly
indicated in these tests that the network itself was not
the bottleneck, but instead the S4Cplus controller. If
the emScon server were to send data more times per
second, another issue would appear in the integration;
Robots have their own controller with ramp-ups and
brake patterns that will slow down the actual step-
response time. For rapid update of the robot
manipulator another method would be required, such
as through shared memory communication directly
with the robot controller.
4.4 Programming method
All robot activities were programmed in the Offline
simulation/programming package V5Robotics/R12
from DELMIA. Since the robot in this research does
not operate like normal robots, the post-processor
phase into RAPID code was not done in the OLP-
module as normal. Instead the XML-code prior to this
phase after all simulations and was downloaded to the
Server PC of the metrology-integration and
interpreted there. This enabled the programming to be
performed on a more task-oriented approach. On the
robot existed only one program that was generic and
never changed. The Server PC executed different sub-
routines such as docking, drilling, fixture
configuration etc. The program execution on the
Server PC manipulated, through WebWare, different
PERS (global) variables, reading and writing I/Os etc.
on the robot. WebWare does not support direct
manipulation of move commands. The generic
program concept enabled this to be overridden.
5 CALIBRATION METHODS
Before starting the execution of the metrology
integration, the robot and its additional accessories
needs to be calibrated. This section is presenting
these methods. Figure 5 shows the transformations
between the different coordinate systems of the
integration.
Tr Tr
Re fl T
Tr Re fl
Rob T Rob
T TCP 0T
TCP 0
Refl
TCP0
Rob Re fl
T
Chuck
Figure 5: The coordinate systems and transforms
In the method presented in this section the notation
A
T for the 4 x 4 homogeneous transform is used to
B
describe coordinate systems and transformations (2).
For example, the homogeneous
A
transformation BT describes the frame { B } relative
to the frame { A }, which when talking about
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004
transforms is interpreted as the coordinate { B } in the
{ A } coordinate system.
5.1 Robot chuck
The first step in the calibration was to calibrate the
Re fl
transformation Chuck T , which is where the chuck TCP
is positioned relative to the 6D-Reflector sitting next
to the chuck. Calibration of the Capto TCP is most
sensitive to avoid tolerance build-up, which results in
low absolute accuracy. The method is summarised as
the following:
1. Install the 6D-Reflector in the chuck using the
male Capto part on the 6D-Reflector
2. Measure the 6D-location A
3. Install the 6D-Reflector in one of the
attachments beside the chuck
4. Measure the 6D-location B
It is important to keep the metrology end-effector
absolutely still during measurement. It is
recommended to remove the metrology/chuck end-
effector and rigidly attach it in a bench during this
calibration phase. When the chuck position is given
from the tracker, this calibration is the only thing that
will affect tolerance build-up. The transformation
matrix is given from this equation:
Re fl
T = TrAT TrBT
Chuck
Tr in the equation is the tracker coordinate system.
The 6D-Reflector unit itself comprising the housing
and the male/female Capto was calibrated by Leica
with an accuracy of 10 m.
5.2 Robot base and TCP0
TCP0 (TCP zero) is the default Tool Center Point on
the robot. The transformation from TCP0 to 6D-
Reflector is important in order to define new TCPs
for end-effectors. This method is based on moving
the robot to different configurations, which will
supply data in a large equation system. The equation
to solve is:
T TCP TCP 0
0 T Re fl T = Re flT
Tr Rob Tr
Rob
This equation is easy to follow by looking in figure 5.
Rob
T is achieved by reading values from the robot
TCP0
controller. For each measurement taken from the
Tr
robot controller a measurement for
Re fl
Rob
from the tracker, hence TCP0 T and
measured in pairs. After calibration run
Tr
T are constructed out of variables in an
Re fl
optimisation algorithm implemented the software
package MATLAB Optimisation Toolbox. The
optimisation is performed by minimizing the cost
function:
f = Tr
TRob Rob TTCP 0,i TCP0 TRe fl TrTRe fl ,i
n steps:
During tests, the value of the cost function was low,
and test runs indicated that the robot co-ordinate
system aligned with tracker coordinate system. It is
important to realize that this calibration will not affect
the final positional accuracy of the robot, but merely
the time to reach a high-accuracy position, since the
servo loop on the server PC will continue iterate until
a position is reached within the specified tolerance
interval.
Another method of calibration was also tested which
again moves the robot to different configurations.
The kinematics equation to be solved is:
Re fl 0 TCP 0 Re fl 1
TCP 0 T TCP
TCP 0 T Re fl 2 T = Tr T Re fl 2 T
Tr
Where Refl1 and Refl2 refer to the 6D-Reflector
reading from two different robot positions.
TCP 0
TCP 0T refers to a robot move in end-effector co-
Re fl
ordinates. The transform TCP 0T is built up from a
series of robot moves using the Powell optimisation
routine.
5.3 End-effector calibration
When manipulating a robot with end-effectors, the
TCP of the robot is often moved from TCP0 to the
end-effector tip. This enables the robot to rotate
around the tip of the end-effector unit. This is
important for reaching high accuracy in the angular
domain. The question here is obvious: How is the tip
of the end-effector calibrated? Note that the
transformation presented in this section will be
presented relative the Robot Chuck. This is because
the implementation presented in this research is based
on attaching end-effectors to the Capto Chuck, not
directly on TCP0. There are basically two different
methods that can be used. One of the methods is
selected depending much on the production scenario.
Inline-calibration of End-effectors
This method is based on the scenario where the end-
effector is already attached to the Robot Chuck and
calibration of the end-effector TCP is to be calibrated.
In this case, the robot obviously needs to be still, and
the production process is stopped. This scenario is
realistic especially if for instance the offset of a
pressure foot is changed; hence the local Z-direction
is moved. In this research, this method has only been
verified for a Drill End-effector; hence the rotation of
the XY-plane around Z was not considered.
T is taken
Tr
T are
Re fl Tr Prism
Rob Bushing T
T and Drill
TCP0
Tr Re fl bushing
BushingT
Tr
T Refl Cylindrical
Re fl object
Figure 6: Calibrating an end-effector
The method described according to the following
 Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004
1. XY-plane: Measure the plane of the pressure-
foot using a reflector ball
2. Z-direction: Measure a cylindrical object in the
same direction as the drill on the machine. In
this case it is not important that the cylindrical
center is align with the drill Z-axis.
3. Z-center: Position the reflector ball in the drill
bushing. The reflector ball does not need to be
on the XY-plane. The measured point is simply
transformed to the XY-plane.
The transform cannot be measured directly but
Tr
T and Tr
T can be measured. Then
Bushing Re fl
is calculated according to:
Re fl
Bushing T= ( Tr
T
Re fl )
1
Bushing
Tr
T (3). Helin P., Jerregrd H., Robertson A., and Snell J.,
Offline-calibration of End-effectors
Probably the most reliable and obvious method to
calibrate the TCP of an end-effector is to measure the
unit in a CMM (Coordinate Measurement Machine).
A CMM has an absolute accuracy in the order of 10
m, hence presumably better than a hand-held
reflector ball. Basically the same transforms as the
Inline-calibration method can be used. The difference
is that the probe of the CMM will measure the Capto
interface of the end-effector. Since this Capto
interface is where the robot holds the end-effector,
the transform to solve in the Offline-calibration
Chuck
method is Bushing T . This transform is then feed in the
robot controller as the new tool. The Off-line method
is more time consuming than the In-line method, but
gives higher accuracy and a more stable environment.
Selecting one of the presented methods must be
decided depending on the production scenario. More
time available and higher requirements on accuracy,
then the Off-line method is appropriate. In some
cases however, moving the offset of a pressure-foot,
would be a waste of time to measure in a CMM,
when the metrology system is already available, if the
accuracy from the metrology system is enough that is.
It is a good role of thumb to calibrate in much higher
accuracy than what is needed. Tolerance build-up is
always raising fast.
6 CONCLUSION
This paper presented continuing research on
Metrology-integrated Industrial Robots. The result
from tests and calibration showed that Industrial
Robots can reach extreme absolute accuracy down to
+/-50 m if a metrology system is online with the
robot controller. Using TCP/IP cannot guarantee real-
time control. In order to reach higher productivity, a
platform that supports real-time control would be the
natural way to continue this research.
7 ACKNOWLEDGEMENTS
This work is part of the EU-founded project
ADFAST (Automation for Drilling, Fastening,
Assembly System and Tooling). Special thanks goes
to the project manager of the ADFAST project, John
Andersson. We would also like to thank SANDVIK
AB for showing great interest and providing to us the
Capto System.
8 REFERENCES
(1). DeVlieg R., Sitton K., Feikert E. and Inman J.,
ONCE (ONe.sided Cell End effector) Robotic Drilling
System, SAE Aerospace Automated Fastening
Conference & Exposition, October 1-3, 2002
Re fl
T (2). Craig J. J., Introduction to Robotics Mechanics
Bushing
and Control, Addison-Wesley, ISBN-0-201-09528-9,
1989
Technologies that make a robot reach Absolut
Accuracy, Proceedings of the 33rd International
Symposium on Robotics, October 7-11, 2002
(4). Kihlman H., Affordable Reconfigurable Assembly
Tooling - An Aircraft Development and
Manufacturing Perspective, Licentiate Thesis No.
980, LiU-TEK-LIC-2002:53, Department of
Mechanical Engineering, Linkpings Universitet, SE-
581 83 Linkping, Sweden, ISBN 91-7373-460-8,
2002
(5). Kihlman H. and Engstrm H., Affordable
Reconfigurable Tooling, SAE 2002 Transactions
Journal of Aerospace, ISBN: 0-7680-1285-6, 2002
(6). Kihlman H. and Loser R., 6DOF Metrology-
integrated Robot Control, Aerospace Automated
Fastening Conference & Exhibition (Aerofast);
September 8-12; Palais des Congrs; Montreal,
Quebec; Canada, 2003
(7). Kihlman H., Eriksson I. and Ennis M., Robotic
Orbital Drilling of Structures for Aerospace
Applications, SAE Aerospace Automated Fastening
Conference & Exposition, October. 1-3, 2002
(8). Kyle S., Loser R., Warren D., Automated part
positioning with the laser tracker, Fifth International
Workshop on Accelerator Alignment, Oct 13-17,
1997
(9). Loser, R., The Authority of Common Measuring
Task Through the Integration of ADM with Laser
Tracker, Coordinate Measurement Systems
Committee, Williamsburg Hospitality House,
Williamsburg, Virginia, July 8 12, 1996
 Appended Publication VII

Kihlman, H., Ossbahr, G., Engstrm, M., Anderson, J., (2004) "Low-cost Automation for
Aircraft Assembly", 2004 Aerospace Manufacturing and Automated Fastening
Conference & Exhibition, September 20-23, Sheraton West Port Hotel, St. Louis,
Missouri

Low-cost Automation for Aircraft Assembly
Department of Mechanical Engineering, Linkping University
Copyright 2004 SAE International
ABSTRACT
In this paper solution for low-cost automation of aircraft
assembly is presented. The concept of this development
is closely related to Lean Automation, which in this
case concerns the use of modern standard equipment
such as standard robots, PC-computers and a newly
developed spatial sensor system for precision
measurements of positions. The robot is used to
perform reconfiguration of tooling modules that are
possible to be configured/reconfigured in six degrees of
freedom. A prototype developed as the result of an EU-
project called ADFAST* has been evaluated at
Linkping University in Sweden. Technical functionality
is reported where the robot manages to configure the
flexible tooling modules to a total error bellow 50 m.
This paper presents the results on the portion of the
project addressing robot, metrology system and tooling.
INTRODUCTION
Today, the majority of aircraft manufacturers use
Dedicated Tooling and other specialized equipment for
the assembly of aircraft structures. The concept of
Dedicated Tooling entails welding beam together into a
framework, whereby rods and pick-ups are attached on
the beams to secure the datum points for the parts to be
assembled. These rods and pick-ups are tailored for
each application and are not re-used. One advantage of
this method is that it is possible to assemble all aircraft
structures, no matter what the level of complexity. The
continued setting up of product components is normally
performed manually, as are subsequent drilling and
riveting operations. Also in the latter operations,
dedicated tools, such as drilling templates, are used.
In the changeover between one product and another,
the components in Dedicated Tooling are not re-used
for the next version of tools. Implementing changes in
2004-01-2830
Henrik Kihlman, Gilbert Ossbahr
Magnus Engstrm
SAAB Aerostructures
John Anderson
Advanced Technology Centre, BAE SYSTEMS
existing tooling requires re-design, machining and
welding of new pick-ups. Thus building a new setup of
tooling can require lead-times up to six months and
sometimes more. An important question is this: how can
lead-times be shortened in changeovers, and tooling
systems be re-used to save money?
One approach could be to have the ability to either re-
build a tooling system, such as the well known Modular
Tooling method [1] that uses extruded aluminum
profiles, where beams are easy to attach/detach. A
modular framework together with 6DOF pick-up
mechanisms combined with metrology has proven to
reduce lead-time for changeovers.
Another method to reduce changeover lead-time even
more is having servos integrated in the pick-ups, where
the tool more or less becomes a CNC-machine. The
latter method becomes problematic due to the
complexity of aircraft products that tends to result in
machines for automated aircraft assembly that are
likewise complex and typically specialized. The
increased assembly speed in the automated system will,
in combination with the fact that the total number of
each airplane model being built is often low or medium
sized, results in a low utilization grade. Most of the time
the assembly machines are standing still, like big
immovable monuments; the nickname Monumental
Automation denotes this kind of automation. The
opposite to Monumental Automation is Lean
Automation. This principle of automation is
characterized by ideal flexibility, simplicity and low-cost
automation. With Lean Automation all Dedicated
Tooling is replaced by Flexible Tooling, and all
specialized equipment is replaced by flexible standard
equipment such as robots, PC-computers etc.
This paper is presents a summary of a European
approach to new methods and techniques in the
building of aircrafts. The consortium consisted of five
major European aircraft manufacturers, four suppliers
and one university. The red thread in this approach has
been the study of how anthropomorphic robots can be
used for automation of drilling, fastening and tooling
configuration. Industrial robots today have many
limitations, especially in stiffness and accuracy. They
are, however, much cheaper than other mechanized
systems currently available for automation. Previous
papers from this research project have presented a new
drilling technology, called Orbital Drilling [2], [3], [4],
Robot-manipulated Tooling [1], [5], and on Metrology-
integrated Robot Control [6], [7]. This paper summarizes
the results from a demonstrator that has been built and
evaluated at Linkping University. The demonstrator
prototype is a robot-manipulated tooling system which
includes robot-integrated metrology to achieve
maintained absolute accuracy in serial-linked robots.
THIS APPROACH TO LEAN AUTOMATION
The solution presented in this research is positioned
somewhere between the previously mentioned methods:
Modular Tooling and CNC-Reconfigurable Tooling, with
respect to how often the tool is reconfigured and the
level of geometric flexibility. In earlier published papers
from this research, this method went under the acronym
ART (Affordable Reconfigurable Tooling), as shown in
figure 1. The characteristic of ART is that a normal
industrial robot is used to configure and re-configure the
tooling actuators to the spatial positions in the 6DOF
needed to give the components of each product their
specific localization during the assembly process. Note
that the actuators in ART are passive, hence they are
only able to maintain a configuration by using locking
devices, and thus not built-in servos or encoders are
used, and all the movement is performed by the robot.
ART is able to re-build similar to Modular Tooling and
has the ability to re-reconfigure similar to CNC-
Reconfigurable Tooling. In the ART concept the
actuators are called Dynamic Modules and the
framework holding the Dynamic Modules in defined
positions simply the Static Framework.
Figure 1: An overview of the ART Tool
Using the Robot to configure the Dynamic Modules in
6DOF has the advantage that the accuracy in the Static
Framework does not have to be highly calibrated as in
CNC-Reconfigurable Tooling. The base plate of each
Dynamic Module needs only to be positioned within a
couple of centimeters. This not only simplifies the re-
build process and setting up of Dynamic Modules on the
Static Framework, but it also allows the system to be
partly configured manually, which results in a low level of
complexity in the system. The idea is to rely on the
robot to position the datums within +/-0,2 mm, which
today is the de facto standard for datum positioning in
aircraft manufacturing. This accuracy requirement
certainly is not always the case however. Accuracy
requirements are sometimes less demanding and in
some cases tougher, such as hinge lines. A normal
industrial robot today has a positional (absolute)
accuracy of around 1 mm. If a good model is calibrated
on these robots, this can come down to 0,5 mm and
even less for smaller robots [8]. This accuracy is only
the case in what is called a non-contact scenario, which
is sufficient in pick-and-place operations. In aircraft
automation however, operations are in most cases
contact-scenarios, hence we cannot rely on the
calibrated model of the robot. Research to solve this
problem, by modeling the environment, has been
undertaken [9]. Modeling the environment however, is
not an easy task. In the case of reconfiguring Dynamic
Modules, friction and load are variable depending on
many factors. This research has focused on supervising
the Industrial Robot with an external control-loop using a
6DOF metrology system.
In addition, this research uses a method similar to the
simulation-based approach [10]. Having a flexible
assembly system in production, must not lead to moving
lead-time from the workshop floor to the office, hence
this technology involves methods for simplifying
operation planning. State-of-the-art simulation systems
today are brought forth for the automotive industry
where the products are built in the range of hundreds of
thousands; in the aerospace industry, the numbers
rarely exceed a couple of thousands. In operation
planning for car assembly, there are times available to
make a good simulation approach. When simulating and
programming robots in the range of 20 products, which
could indeed be the case when working with a flexible
assembly system, the resource to simulate is much
more limited.
Essentially the approach to low-cost automation of
aircraft assembly, according to the indicated principles
of lean automation, requires that:
Dedicated tooling of all kinds should be either
eliminated or replaced by flexible and re-
programmable equipment
 The efforts for programming of the system and
changeover between product must be low
The system complexity should be lowered
compared to what is reached in conventional
automation
The flexibility in relation to product variability
should be big enough to enable different
products to share the available manufacturing
capacity thereby enabling a good utilization
factor
The next two sections will present in more detail what
constitutes the ART system.
THE STATIC FRAMEWORK DESIGN
The Static Framework of an ART has the task to keep
the Dynamic Modules of the system in firm position
during the assembly process. The demands on stiffness
and stability of the framework are generally high and are
of the same level as for conventional assembly tools.
The demands on accuracy, however, are essentially
lower, due to the fact that the fine-tuning of the pick-up
point positioning of an ART is made by the robot and
consequently is decided by the robots accuracy rather
than that of the framework. This is a fact simplifying the
concept of ART compared to conventional tooling, and
is a contributing factor in its affordability.
A survey of the market, however, failed to find any
commercially available, ideal construction kits for this
purpose. Either the stability of the components was too
low, which typically was the case for aluminum profile
systems, or the costs were too high, as was the case for
machine tool fundament systems. As a result, a new
modular static framework construction kit was specified
according to the requirements and developed within the
project. The preliminary name of this new design was
Box-joint.
The new Box-joint construction kit uses standard steel
beams with square or rectangular cross-sections in a
modular series of measures. This is to enable the
required stability at the lowest possible cost. The joints
of the system press the beams together by means of
bolts and pressure plates to keep the beams using
friction, in specified positions. Different sizes of beams
can be combined and moved to feasible perpendicular
positions relative to each other as is illustrated in figure
2

At the start of the project to develop an automated low

cost system for aircraft assembly, it was identified that a
construction kit using movable joints would be ideal for
the construction of the static framework. Using such a
kit, it was though that it would be possible, to quickly
erect the needed fundaments for the Dynamic Modules
Figure 2: A framework design by means of the
with respect to the limited reach of each module. By
developed Box-joint system
means of a correct analysis of the spatial needs defined
by the different products to be assembled in the tool, it
The pressure plates belonging to the system are also
would be possible to find a placement for each Dynamic
used to attach the dynamic modules in feasible
Module, which, in the majority of cases, could be
positions along the beams of a framework as is
common for the different products.
indicated in figure 3. Thus, varied functions of the
framework are enabled by means of a minimum number
In cases where the reach for a certain Dynamic Module
of different construction modules.
was too short, or when quite another placement was
needed, this desire also could be satisfied by means of
By means of calculations and experimental tests the
a construction kit for the ART-framework. The module
applicability and stability of the developed construction
then would have to be manually loosened and moved to
kit for static framework has been verified.
the new position a manual effort what seems
reasonable due to the low accuracy claims, facilitating
Further more, the use of air-cushions for the transport of
this kind of work.
the static framework on a floor is recommended. This
method of enabling larger workspace of the robot
Also, the introduction of new products in an existing cell
relative to the product, has successfully been tested as
of affordable automated assembly of aircraft structures
part of the experimental work with the system.
would require a construction kit for the static framework.
The tool could then be easily rebuilt to fit the new
product as well.
THE DYNAMIC MODULES
This section presents the Dynamic Modules. Some
parts of the sub-systems presented in this section have
previously been presented in [1]. The Dynamic Modules
were designed so that an operator could attach them to
the static framework manually. Seven different Dynamic
Modules were developed, see figure 3.
Figure 3: The Dynamic Modules
In a normal case there would probably be only one or
two Dynamic Module solutions chosen in an ART
system. In this case, there was seven Dynamic Modules
developed for being able to evaluate as many different
solutions as was possible with respect to project time
and budget. In general a Dynamic Module consists of a
base plate and a top plate. In between there can be a
parallel mechanical structure [11] or a serial linked
structure. The Octapod, Tripod and Hexapod are
examples of parallel mechanical solutions, and the
Cradles are examples of the serial linked solution. The
Carriage solution is a hybrid between serial- and parallel
structure. The base plate is attached on the framework,
and has therefore the same design as the Box-joints
and is attached accordingly.
LOCKING MECHANISM
The locking sleeves are hydro-mechanical, which
means hydraulic pressure in the system is only required
when changing from unlocked to locked mode and vice
versa. The hydraulics pushes a wedge between two
states and thereby activates the mechanical locking
effect. This movement is crucial so as not to affect the
position of the legs in the Dynamic Module. Using the
Hypus locking sleeves from the company ETP
Transmission, made this possible.
This research has focused on using cheap off-the shelf
components, preferably mass-produced for other
purposes. This was however, not always possible, and
some sub-components had to be developed in-house,
such as the locking mechanisms for the ball joints.
THE CAPTO INTERFACE SYSTEM
At the top of the Dynamic Modules plate is attached an
interface called a SANDVIK Capto. The Capto interface
has several purposes, see figure 4. A primary function is
for the robot to dock onto the modules. A secondary
function is to attach the pick-ups that are holding the
aircraft parts. A tertian purpose is for the Metrology
probe to attach to the modules to calibrate their initial
position. The Capto system is robust and has a
repetitive accuracy of 2 m and was originally designed
for holding cutting tools in CNC machines, hence they
are mass-produced and relatively cheap. The robot, the
pick-ups and the Metrology Probe have a female Capto
chuck. The robot chuck is locked automatically, whereas
the probe and pick-up chucks are locked manually.
Figure 4: Three applications for the Capto System
THE PICK-UPS
After the robot has positioned the Dynamic Modules to
an accuracy of 50 m, the pick-ups are attached on the
top-plate of a Dynamic Module. The important accuracy
however is the datum point. A datum point is the
position in Cartesian space that defines where the
aircraft parts are to be located, or picked up, see
figure 5. One issue is the fact that the position given by
the robot during configuration of the Dynamic Module,
ensures accuracy of the Capto interface and not the
datum point.
DMs positioned by
the Robot
Datum point
Figure 5: The aircraft parts are attached on the
datum points
Figure 5 shows the relation between the male Capto from ABB has a similar system called WebWare. Using
interface on the Dynamic Module (DM) that is position emScon and WebWare, the tracker and robot are
by the robot and the datum point that is attaching the integrated by a Visual Basic program. In its current
aircraft part. This relation was calibrated using a state, the integration is not a real-time system, and it
Coordinate Measurement Machine (CMM). From the communicates on the level of seconds. The topology of
measurements the CAD model of the Pick-up was the system is shown in figure 6.
updated. The updated value is in fact the difference
between the nominal configuration and the measured
relation. As a result the male Capto position in Cartesian
space was updated in the simulation model, and the DM
was re-positioned slightly, to compensate for the
difference between the digital model of the Pick-up and
the physical one. This method gave the datum points
good accuracy.

DUMMIES

Five of the eight DMs were locked hydro mechanically.

The other three was locked manually and was called
Dummies. There were only enough resources to build
one active Dynamic Module of each kind. The Dummies
proved to work surprisingly well. Even though the
Dummies were locked and un-locked manually, they
where stilled configured by the robot similar to the active
ones. In fact the same reconfiguration cycle as the
hydro-mechanical solution was performed, but with
manual lock/un-lock operations.
Having presented the building blocks of the ART
system, the next section in this paper will go deeper into
the integration of robots and metrology system that
enables maintained accuracy in reconfiguration of the
DMs.
THE METROLOGY-INTEGRATED ROBOT
CONTROL
Figure 6: Topology over the Integration
Experiments proved that it was possible to position the
robot to within 50 m (square root sum of the three axis
error). This interval could easily be changed. Several
tests were done on 25 m, but that sometimes caused
oscillating behavior, since 2-3 m away from the tracker
unit is the accuracy limit of the tracker unit. Normal use
of the LTD800 is to use it for probing. The handheld
probe is called the T-Probe. The T-Probe is much more
convenient for probing instead of having to work with a
small prism ball. In this research an additional Robot
Reflector was developed, see figure 7.

Industrial robots today do not meet the accuracy

requirements for aircraft automation. However, they are
mass-produced primary for the automotive industry and
are therefore much cheaper than other machines for
automation, hence solving the inaccuracies would give
us a low-cost machine for aircraft automation. One way
to compensate for the lack of accuracy is to integrate
the robot controller with a metrology system. This
research has used the LTD800 from Leica Geosystems.
The LTD800 laser tracker measures through an
interferometric laser the distance to a laser prism.
Positional accuracy is 10 m per meter distance away
from the tracker unit. The LTD800 is the next version of
the LTD500 with an additional camera called the T-
Cam. The camera measures orientation of a probe,
using 10 LEDs. The T-Cam has a built in zoom lens, Figure 7: A reflector attached on the robot
which is used into zoom in the probe to fill the image
prior to processing it. The T-Cam gives an orientation The Robot Reflector was designed using the same
accuracy of 0,02 degrees throughout the work volume of Capto system as for the Dynamic Modules. It contained
15 m. The LTD800 is possible to be controlled over a 10 LEDs and it works the same way as the T-Probe.
normal Ethernet LAN using emScon, an embedded Using the Capto system for the Robot Reflector enabled
system control software. The IRB4400 Industrial robot probing of the initial state of the Dynamic Modules.

Figure 8: The coordinate systems and transforms
Reconfiguring a Dynamic Module involves friction. As
mentioned earlier, modeling your environment is one
approach to dealing with the stiffness and accuracy
problems of articulated robots. This research, solving it
by supervising inaccuracies, was successful, and the
extra time to manipulate the Dynamic Modules, moving
with friction, was just slightly longer. The calibrations
and transforms, as shown in figure 8, is more in depth
presented in [6] and [7].
OPERATION PLANNING
This paper so far has presented and discussed one
philosophy to enable low-cost automation for aircraft
assembly. We want to use industrial robots that are
normally used in automotive industry, and although
these robots do not have sufficient accuracy, they still
are relatively cheap compared to other machines used
in automation.
Todays Industrial Robots are programmed by off-line
systems. Although programming has now become
object oriented rather than robot oriented, which
essentially means you need not understand a specific
robot language and programming is done by clicking on
objects in a 3D graphical environment. But, comparing
the different scenario for the manufacturing of cars with
aircrafts; cars are built on the range of hundreds of
thousands and aircrafts may be built in the range of a
few thousands and often fewer. The simulation- and
programming systems for Industrial Robots available
today was brought about for car industry. In the scenario
of building hundreds of thousands of cars, there is more
time available for operation planning compared to
building just a few aircrafts. Programming and operation
planning must be simplified in a low-volume production
scenario, such as for aerospace.
This research has included studies to simplify
programming of robots by moving up towards the
process-oriented abstraction level using V5Robotics
from DELMIA. A process naming of the frames (robot
locations) was developed. DELMIA and many other
simulation systems use the standard language XML.
Normally XML is stored in the background of a
simulation and not until a specific robot language is
selected in the OLP (offline programming) module of the
simulation system, is the XML then transformed
(parsed) to a robot specific code. In this research, the
parsing was never performed in the simulation software.
Instead the XML code was downloaded to the
integration software in the workshop. The integration
software interpreted the processes from the XML
directly, hence parsing it to robot operations, operations
for the operator, measurement tasks for the tracker etc..
It is important to realize that this system is not just a
robot cell. It includes several tasks that are not robot
specific. This method also allowed the operator to
execute any sub-process along the sequence of
operations in the task list that was presented exactly the
same in the integration software as in the simulation
environment. This method of programming was possible
to implement by having only one small generic program
in the robot controller, hence no dedicated robot code
for each sequence of operations.
By using a metrology system to guide the robot to high
accuracy, no normal robot calibration of the system was
required when going from the offline-programming
system to the workshop floor. The only calibration
performed was the Probing of Dynamic Modules
presented earlier. This indeed simplified implementation
of changes that was brought back to the simulation
model from physical experiments.
RESULTS
This section will discuss some interesting issues that
arose during the final implementation of the
demonstrator and describe measures undertaken to
solve them.
STATIC FRAMEWORK
The ideas of constructing the demonstrator with
modular beams went better than expected. From the
stage of having beams on the floor until the complete
framework according to figure 1 took one day. That was
accomplished without using a crane. Instead the winch
system on top of each vertical beam was successfully
used. There where some difficulties to model the
mechanical limitations in the digital model of the
Dynamic Modules, which resulted in some late changes
in the framework that easily could be modified thanks to
the ability to re-build.
DYNAMIC MODULES
The Dynamic Modules had some issues in the final
testing. The ball-joints had some axial movements
during locking. Where a wedge is pushing the ball onto
a metallic ring, movements occur during locking. This
could be a problematic case, since the robot ensures 50
m only prior to locking. When locking occurs, the robot
cannot compensate anymore. Two different approaches
to solve this problem were tested. The first one is to lock
each individual locking sleeve in a certain order. If the
ball joint is locked first, the movement in the ball will be
taken up in the rest of the module. The second
approach is to measure the position of the robot after
locking has occurred. This way the error due to locking
is identified. In this case, the module is first un-locked,
and then re-adjusted by the robot, where the error
during locking is added to the first position. When
locking is done again, the module will lock itself to the
right position.
TRANSFORM CALIBRATION
Another issue in the final testing was the calibration of
the metrology system to the robot. To calibrate the
tracker relative to the robot, a sequence is performed
where the robot is moved to several locations with
changes in orientation. Locations are stored both in
tracker and robot and calculated. More information
about this can be found in [7]. Due to the relatively bad
absolute accuracy in the robot, the transformation will
have an error. Look for the dotted lines in figure 10.
When the robot is iterated down to a square root sum
off 50 m, the robot reference values are way of. The x
reference, for example of the robot in figure 10, shows
1.6 mm off.
Figure 10: The robot and tracker position in a high-
accuracy point
It is important to realize that the transformation error
does not affect the absolute accuracy of the integration
itself. The integration is controlled in relation to the
tracker coordinate system, hence what you see is what
you get is the case here - at least as well as the tracker
can provide us regarding accuracy.
REFLECTOR CALIBRATION
As shown in figure 4, the Sandvik Capto interface on the
reflector was used for probing the initial state of the
Dynamic Modules can in addition could be attached on
the robot during metrology-integrated robot control. The
calibration of the female Capto on the probe did not fully
meet the expectation. When docking the robot to the
Dynamic Modules a clearance of 1 mm in the entrance
was required. Due to some misalignment in calibration,
docking in the preliminary testing was for this reason
sometimes not possible. This was later solved by
calculating the difference in the measured value
between probed location and docked location. The
difference was added to the docking position. The result
was indeed satisfying and not one docking was missed
after this point.
UN-DOCKING
The metrology system only guides the robot to its final
positional accuracy at a location, not on the trajectory
between two locations. And due to the rather low robot
accuracy the robot risk applying un-wanted forces
during un-docking. The forces may be applied between
the inner side of the female robot chuck and the male
Capto interface on the DM. As long as this bending
movement is lower than the stiffness limits in the DM,
nothing happens. In some tests the robot did push too
hard and caused a small deviation of the datum point.
This was rather easily solved by moving the robot in its
TCP coordinate system by the ABB RAPID command:
RelTool. RelTool moves linearly in tool coordinates.
Although the robots have bad absolute accuracy, they
are good at moving relatively to a location they have
been in before, in this case the docked position. A
second approach to solve this problem was tested in
addition. This was done by making the robot less rigid
by using something called soft-servo. The disadvantage
of using soft servo is that when the undock sequence is
complete the robot sometimes makes big movements
that are hard to control. Therefore it was better to make
the undock movement as linear as possible to avoid
unwanted forces on the Dynamic Module.
FINAL REMARKS
Striving towards low-cost automation, it is imperative to
use as much existing components as possible instead
of making everything in-house to get it right. The
challenge was to design systems with novel ideas from
what is already available on the market. This could keep
cost down, and quality ensured, which was important in
the hunt for errors and causal effects when putting sub-
systems together. Furthermore, buying existing
components will also save lead-time. Excluding these
facts may lead to one of the major problems in
automatic assembly systems, especially for complex
products such as airplanes, where assembly systems is
common to be vastly expensive and suffer from large
technical risk and are in-flexible. Again, this is what we
call Monumental Automation.
CONCLUSION
This paper presented the final results of the evaluation
and testing of an aircraft assembly demonstrator that
was built at Linkping University as a result of a three-
year project funded by the European Commission. A
robot has successfully reconfigured Dynamic Modules to
50 m mm accuracy. Some modules did move slightly
during module locking, but several methods have been
investigated and tested to deal with this problem.
Although the metrology system used for this research is
somewhat overqualified, it has given us good quality in
metrology feedback necessary to reach aircraft
requirements on positional accuracy. Future research at
Linkping University will investigate new low-cost
solutions for metrology-integrated robot control.
Promising efforts are now being made in also
integrating drilling and fastening functions into a
complete automatic low-cost system.
REFERENCES
1. Kihlman H., Affordable Reconfigurable Assembly
Tooling - An Aircraft Development and
Manufacturing Perspective, Licentiate Thesis No.
980, LiU-TEK-LIC-2002:53, Department of
Mechanical Engineering, Linkpings Universitet,
SE-581 83 Linkping, Sweden, ISBN 91-7373-460-
8, 2002
2. Lindqvist, R., Eriksson, I., Wolf, M., Orbital Drilling
of Sandwich Constructions for Space Applications,
SAE Aerospace Automated Fastening Conference
& Exposition, 2001
3. Kihlman, H., Eriksson, I. and Ennis, M., "Robotic
Orbital Drilling of Structures for Aerospace
Applications", SAE Aerospace Automated Fastening
Conference & Exposition, October. 1-3, 2002
4. Lindqvist, R. and Kihlman, H., Orbital Drilling -
Implementation and Evaluation, SAE Aerospace
Automated Fastening Conference & Exposition,
Sept. 20-23, 2004
5. Kihlman, H. and Engstrm, M., "Affordable
Reconfigurable Tooling", SAE 2002 Transactions
Journal of Aerospace, ISBN: 0-7680-1285-6
6. Kihlman, H. and Loser, R., "6DOF Metrology-
integrated Robot Control", Aerospace Automated
Fastening Conference & Exhibition (Aerofast);
September 8-12; Palais des Congrs; Montreal,
Quebec; Canada, 2003
7. Kihlman, H., Sunnanbo, A., Loser, R., Von Arb, K.,
Cooke, A., "Metrology-integrated Industrial Robots
Calibration, Implementation and Testing", 35th
International Symposium on Robotics, Paris-Nord
Villepinte, France, March 23-26, 2004
8. Helin, P., Jerregrd, H., Robertson, A., and Snell, J.,
"Technologies that make a robot reach Absolute
Accuracy", Proceedings of the 33rd International
Symposium on Robotics, October 7-11, 2002
9. Dgoulange, E., Dachez, P., Pierrot, P., Prat, P.,
"Robust Design of Independent Joint Controllers with
Experimentation", IEEE Transactions on Industrial
Electronics, vol. 40, No. 4, August, 1993
10. Webb, P., Eastwood, S., Chitiu, A., Fayad, C.,
McKeown, C., The Design and Realisation of a
Flexible Rapid Assembly Aero-Structure
Manufacturing Cell A simulation Driven Approach
Aerospace Manufacturing Technology Conference &
Exhibition; September 8-12; Palais des Congrs;
Montreal, Quebec; Canada, 2003
11. Askbrink, P., Optimization of Parallel Mechanical
and Adjustable Assembly-jig, LiTH-IKP-Ing-Ex-
02/14-SE, June 2002
CONTACT
Henrik Kihlman
Department of Mechanical Engineering
Linkping University
Phone (+46) 13-288 974
E-mail: Henrik.Kihlman@ikp.liu.se
Gilbert Ossbahr
Department of Mechanical Engineering
Linkping University
Phone (+46) 13-281 129
E-mail: Gilbert.Ossbahr@ikp.liu.se
Magnus Engstrm
SAAB Aerostructures
Phone (+46) 13-184 585
E-mail: Magnus.Engstrom@saab.se
John Anderson
Advanced Technology Centre
BAE SYSTEMS
E-mail: john.s.anderson@baesystems.com
ABBREVIATIONS
ADFAST Automation for Drilling, Fastening, System
Integration and Tooling
ART Affordable Reconfigurable Tooling
DM Dynamic Module
 Appended Publication VIII

Lindqvist, R, Kihlman, H.,(2004) "Orbital Drilling - Implementation and Evaluation",

2004 Aerospace Manufacturing and Automated Fastening Conference & Exhibition,
September 20-23, Sheraton West Port Hotel, St. Louis, Missouri

Orbital Drilling - Implementation and Evaluation
Department of Mechanical Engineering, Linkping University
Copyright 2004 SAE International
ABSTRACT
This paper deals with issues about Orbital drilling
implementation and evaluation. The paper summarizes
and includes the so far written papers about Orbital
drilling.
In a previous paper, a conclusion was made, orbital
drilling is a suitable method for drilling high quality holes
in the specific space application that was studied. The
conclusion for further development was then to look
more specifically into the process development, i.e.
cutting parameters, new cutters, new coatings on
cutters etc.
Then the investigation and feasibility study started. The
project went from early prestudy to fully implemented
industrialized solution of the PODU at SAAB Ericsson
Space (SES) in Linkping, Sweden. More detailed
description on how the project did it is explained in the
paper.
The early conclusions made after pre- and final
acceptance tests of the PODU at SES shows that
Orbital drilling is a comprehensive method compared to
conventionally drilling methods. Today the Orbital drilling
method is a certified production process at SES.
Another part of this paper deals with issues about
evaluation and the possibility to integrate Orbital drilling
units on standardized industrial robots (IRB). The
questions are why they are of interest and how they can
reduce the extensive use and cost of hard or dedicated
tooling, i.e. drill jigs and templates.
INTRODUCTION
Three and a half years ago the European partly funded
project ADFAST (Automation for Drilling, Fastening,
Assembly, System integration and Tooling) started. One
part of the project was to study manual drilling using the
2004-01-2814
Richard Lindqvist
SAAB Aerostructures
Henrik Kihlman
Portable Drilling Unit (PODU) for different aerospace
materials. In total four end-users participated, i.e. Airbus
UK, Airbus Spain, Alenia and SAAB. Each one of the
end-users was allowed to choose maximum two
applications each. SAAB decided to study two
applications:
New future application with a stack of materials
comprising of: 10mm Aluminium + 20mm
Carbon Fibre Reinforced Composite (CFRP) +
10mm Aluminium.
SAAB Ericsson Space application with a stack
of materials comprising of: 0,8mm CFRP +
18mm Aluminium Honeycombcore + 0,8mm
CFRP
The first part of this paper will present results from the
tests performed in these applications using a PODU
TwinSpinTM 201 that was designed and developed by
Novator AB for SAAB Ericsson Space in Linkping,
Sweden.
The PODU used at SAAB Ericsson Space was a later
version from the predecessor, see figure 1 that was
studied in parallel in the beginning of the ADFAST at
SAAB Aerostructures. The unit in figure 1 is the first
known developed and manufactured version of the
PODU.
Figure 1. The first developed and manufactured
PODU with drill jig, used for drilling trials
In addition to the drilling experiments of the PODU, a No delamination from the drilling process is allowed in
cost-model was developed in the ADFAST project and the CFC parts. Fiber fraying is allowed in one layer with
presented by Mr. Latger et al. at the SAE 2002 Aerofast the size of 1x2 [mm] in squares.
conference [1]. This model has been used in order to
evaluate Conventional drilling methods over Orbital The first drilling trials were carried out and were
drilling methods. The output from the cost-model has presented by Lindqvist, et al. at the Aerofast 2001
acted as input facts for business decision regarding conference [3].
investment in Orbital drilling Technology at SAAB
Ericsson Space. INVESTIGATION PHASE

Another result from the ADFAST project has been the After the prestudy and on the behalf of the ADFAST
results from the fatigue test studies in CFRP presented project, the parallel driven internal SAAB Aerostructure
by Mr. Harris et al. at the Aerofast 2002 Conference [2]. project was set up and launched. Novator AB became
The main conclusion was then was Orbital drilling in more closed engaged in the project and decided to set
CFRP coupons indicated longer fatigue life and that the up a test program for SAAB.
static strength properties are equal to Conventional
drilling. These results presented in that paper was Development of the PODU
important for the continuation and implementation of
Orbital drilling technology, i.e. implementation and The original ADFAST developed PODU was designed
evaluation at SAAB Ericsson Space. with a 4 mm offset, which in theory would allow a
maximum hole diameter of e.g. 12 mm to be drilled with
As a result from the ADFAST project was also the study an 8 mm cutting tool, if the characteristics and benefits
of automating the Orbital drilling process. The second of Orbital drilling were to be realized (cutting tool to hole
part of this paper, in addition to manual drilling at SAAB diameter ratio).
Ericsson Space, there will be a presentation of results
from Orbital drilling using a standard articulated robot. A
A second PODU with a 12 mm offset was made
demonstrator was developed to investigate if it was
available to meet the specified diameters in the SES
possible to put a CNC Orbital Drilling Unit (CNC-ODU),
application. It became apparent, at an early stage, that
the TwinSpinTM 102 from Novator AB, on a KUKA KR
200 Robot. This section will start with a general this was the only logical way to meet these
discussion on using standard, light-weight robots for specifications, if the characteristics or benefits of Orbital
automation of drilling for aerospace materials. drilling were to be realized,

This modification involved designing and building a

IMPLEMENTATION AND EVALUATION OF PODU
second PODU with a 12 mm offset so that a 22 mm
TWINSPINTM 201 AT SAAB ERICSSON SPACE
cutting tool could be used to drill a 33 mm hole. In other
words the PODU with a 12 mm offset could be offset in
This first part of the paper is based on the experiences
excess of 11 mm to allow up to 33 mm holes to be
from the evaluation and implementation project of the
drilled with a 22 mm cutting tool in the SES application.
Portable Orbital Drilling Unit TwinSpinTM 201. In the
ADFAST project SAAB Ericsson Space Cylinder
Cutting tools
selected application, see figure 2, was studied because
of the SAAB Ericsson Space (SES) early interest of
The cutting tool design used in CFC applications was
industrialization of Orbital drilling technology in their
advanced considerably from its original concept at the
potential Space program, i.e. the future Space Cylinder
start of the ADFAST project. The original design used
production to the customer EADS Alcatel.
PCD tipped carbide cutting tools, see figure 3. These
tools represented the status of Orbital drilling PCD tools
PRESTUDY
at the time, and were limited in that they incorporated a
neutral or zero rake angle (or cutting angle) and had no
In the prestudy the specific hole requirements on the
provision for air thorough the cutting tool.
SES application were set, se figure 2 below:

Survey Aircraft Application Material Total Hole sizes Fastener Hole Hole Hole
thickness & type diameter surface alignment
tolerances range roughness
Saab 8 Saab Space C (0.8mm) 19.6mm 12mm up plain 12 up to 6.3 um 90 1.7
Ericsson cylinders +A to hole 33mm Ra in
Space honeycomb 33mm carbon
core (H11)
(18mm) +
C (0.8mm)

Figure 2: The SAAB Ericsson Space application and

requirements

Figure 3. The original design used PCD tipped
carbide cutting tools.
A further advance of this concept, se figure 4, was
introduced into these drilling trials. This tool had two
brazed PCD cutting tip inserts, a 6 rake angle (or
cutting angle) despite the fact that it is brazed PCD, and
had air through the tool shank. Consequently this cutting
tool worked quite well in these tests.
Figure 4: A further advance of the first concept was
introduced
Ironically the efforts to find an optimal cutting tool
solution for these applications resulted in the adoption of
readily available and relatively inexpensive diamond
mandrels, see figure 5; rather than the appreciably more
expensive custom built PCD tooling, used throughout
the cutting tool development. The diamond impregnated
mandrels machined these holes in these materials with
excellent results, and were subsequently incorporated
into the production solution at SES, subsequently
addressed by this technology.
Figure 5. More inexpensive standard diamond
mandrels became the final solution.
So the solution adopted, as stated earlier, was to
orbitally drill, or grind the holes with readily available and
more inexpensive diamond impregnated mandrels.
Extensive tests were carried out using the more
conventional PCD platform, and although these cutting
tools worked well, they were additionally, relatively
expensive. The diamond impregnated mandrels on the
other hand were appreciably less expensive and worked
very well addressing the delamination concerns, dry
drilling, tool wear and provision for debris and dust
removal.
Hole drilling procedure
These holes in the SES application were drilled in a one
pass cycle. When orbitally drilling holes within
aerospace tolerances the offset figure used in
calculating a desired hole diameter with a given tool is
always subjective. In other words it not simply a matter
of subtracting the cutting tool diameter from the final
desired hole diameter, and then entering the resulting
value into the machining parameters, to achieve final
hole diameter. This factor must be assumed, calculated
and applied to the application at hand. This is quite a
simple procedure in practice and involves setting the
diameter to be drilled before commencing with the hole
machining operation. It could be compared to qualifying
a hole or a series of holes in a test or offer sample
before engaging in the actual machining operation,
common in Conventional drilling operations.
Hole calibration
It is important to understand that the Orbital drilling
diameter offset principal allows the operator to choose
a diameter as a starting point for the actual hole
diameter. Obviously this is carried out as a rough
starting point, and when this first hole or set of holes are
measured the operator simply adds or subtracts the
difference from the desired nominal size and re-drills the
hole at this new diameter. In practice this means drilling
a number of holes in a drill-in procedure, especially in
this drilling trial, where cutting tools and parameters are
been tested in this particular CFC/Al/CFC
configuration for the first time.
It is important to understand that it is normally only
necessary to drill a small number (2 or 3) of drill-in
holes whether the intended drilling trial hole is tens or
hundreds of holes. This drill-in calibration procedure
was carried out in drilling jigs built for this purpose, see
figures 6 and 7.

Figure 8. An overview picture of the Bread Board

being fixed in the main jig with a few of the drill jigs
mounted in the main jig.

Figure 6. Calibration drill jig, especially designed

and manufactured by SES.

Figure 7. Calibration drilling in order to set and

calibrate the machine parameters.

Drilling trial results

The results achieved in this particular investigation

study shows evidently it was possible to drill all the holes
within tolerance as specified by SES, at time of drilling.
It is suffice however to state that; the holes drilled in the
SES sample material were within tolerance, given that
the actual acceptance test, in the execution phase of
Figure 9. A more close picture on Bread Board (BB),
the project, encompassed the drilling of 600 holes in the
note the height of the cylinder was only one meter
different hole sizes in the SES Qualification Model QM,
compared to the full size 4 meter high cylinder.
all drilled within tolerance. A large amount of holes were
drilled in the pre-production unit, which were named
Bread Board (BB), see figures 8, 9 and 10.

Figure 10. The PODU mounted in drill jig ready to
drill holes in Bread Board (BB).
EXECUTION PHASE
After the investigation phase the project went in to the
execution phase. In this phase the qualification model
for serial production certification was drilled.
The actual acceptance and pre-production test for
Orbital Drilling, PODU implementation at SES
encompassed the drilling of 600 holes in the different
hole sizes in the SES Qualification Model (QM), all holes
were drilled within tolerance. Below is pictures taken
from the acceptance and pre-production tests at SES
showed, see figure 11 and 12.
Figure 11. An overview picture of the assembly
clean room environment.
Figure 12. Qualification model being drilled at SES.
RESULTS
The SES application has been studied and the PODU
TM
TwinSpin 201 was later implemented during the year
2003 at SAAB Ericsson Space in Linkping, Sweden. In
total, more than approximately 600 holes have been
drilled as a part of the acceptance and preproduction
test program that has been carried out at SES. Only one
hole out of the 600 holes that was being drilled in the
QM was detecting a small and single delamination in
one layer. The main conclusions and results from this
study could be summarized as follows:
Orbital drilling can produce holes within
tolerance without entry, exit or inter-laminar
burrs with PODU machining in these
applications, see figure 13.
Orbital drilling can produce holes with a surface
roughness comparable to ground or lapped
surfaces with PODU machining in these
applications.
No cutting fluid was used. Orbital drilling can
produce holes without cutting fluid with PODU
machining in these applications.
Little dust or debris, this was removed via
vacuum. Orbital drilling can produce holes
without dust or debris with PODU machining in
these applications.
Orbital drilling can produce holes with
substantially reduced axial cutting forces with
PODU machining in these applications.

Figure 13. This picture shows one of the orbital
drilled holes. Note the clean and sharp hole edges.
Today and in this time of writing, SAAB Ericsson Space
performs a static test on the first Qualification Model
(QM). After that the QM model has been qualified the
serial production in full scale will begin.
So far, has been presented, the evaluation of using the
PODU, which is a manually operated drilling process.
Next step in this paper will discuss using articulated
robots for automation of drilling, and in particular by
using the CNC-ODU.
AUTOMATION OF ORBITAL DRILLING USING
ARTICULATED ROBOTS
This section will start out by issues about using
industrial robots for automation of drilling operations.
Conclusions will be drawn on how these robots can be
used if certain measures are taken into accordance.
Before concluding this section on the advantages using
articulated robots for automation of drilling, the
demonstrator being developed by Airbus UK, BAE
SYSTEMS and Novator will be presented.
ROBOTICS FOR DRILLING AUTOMATION
Absolute, or in other words, positional accuracy is in
general +/-1 mm for industrial robots. If well calibrated,
absolute accuracy can come down to +/-0,5 mm or less
for smaller robots [4]. Robot suppliers can today deliver
what is called HA-Robots, or High-Accuracy Robots.
HA-Robots are going through a calibration cycle with a
metrology system prior delivery to customer. This
calibration is mainly done in a computer program, where
a static model is generated that is unique for each robot
being calibrated. The general requirement on accuracy
in aircraft assembly, which is how well parts must be
positioned in relation to each other, is around +/- 0,2
mm. This however can be somewhat confusing when
relating this requirement to drilling. Positioning parts
together is not the same as position holes and rivets.
Some aircraft manufacturers say they want holes for
solid-rivets to be positioned within 2 mm, or an spar
edge distance of 1 mm. Diameter tolerance is required
to stay within 0,2 mm. Surface finish is required to be no
worse than Ra 6,3 m. These values are indeed
general, and in many cases it differs, but may give the
reader a feeling on which accuracy domain aircraft
automation has to cope with. Lockbolt fasteners are
much tougher challenge due to forced fit, which
demands diameter tolerance of 0,05 mm and a Ra of
1,6 m. Normality could be generalized to 1 degree.
This would mean that HA-Robots could in fact manage
the tolerance demands from Aerospace, although with a
tight margin. There is just one problem, however.
Assume that robot manufacturers can guarantee 0,2
mm; that would only be the case if the robot is only
exposed for gravitational loads and that changes in
forces are not too quick. In addition HA-Robots today
are yet only available in smaller versions (30 to 60 kg
payload). In Conventional Drilling axial forces are high
[5]. Forces in drilling are not the same as gravitational
forces, with the drilling unit be attached on the robot
TCP. This indicates that robotic drilling is possible using
light-weight articulated robots if the forces in drilling
were small enough and does not change too rapidly. In
drilling however, forces are changing dynamically. As
presented by Mr. DeVlieg et al. at the SAE 2002
Aerofast conference [6], positional accuracy is hindered
by numerous factors, such as: imperfect kinematic
model, payload droop, misaligned robot mounting,
uneven track rail, flawed TCP definition, imperfect gear
ratios, backlash etc. And even if there where ways to
improve the positional accuracy, there would still be a
remaining issue to deal with the dynamic loads.
One method to overcome dynamic loads is to use a
pressure foot. In this case the robot is using the
pressure-foot to press the end-effector to the surface
prior to engaging the cutter. In this research,
experiments with Conventional Drilling methods has
been tested, and if the pre-pressure is high enough, the
drilling machine will stay normalized during the drilling.
One common problem using light-weight robots, as in
the case mentioned earlier, the robot will loose normality
to the surface. One common mistake is to attach the
drilling unit in a 90-degree angle that will induce this
effect, see figures 14a.
j6
)
j4
Figure: 14a Figure: 14b
Robot suppliers recommend attaching the drilling
machine with the cutter pointing straight out from the
robot TCP (figure 14b). Having the cutter pointing
straight out from the TCP will on the other hand
increase the risk to end up in a singularity, which will
happen when joint 4 and joint 6 comes close to pointing
in the same direction. Joint 4 and 6 is shown also in
figure 14b. In a singularity the robot controller cannot
calculate the inverse (joint angles from Cartesian
coordinates), and the robot will stop. One way to
maintain normality, when applying the pressure-foot,
would be to drill with lower cutting forces.

The Orbital drilling process produces low axial cutting

forces; hence less force needs to be applied by the
pressure-foot. The next part in this section will present a
demonstrator that was developed in the ADFAST
project, constituting a CNC Orbital drilling unit,
TwinSpinTM 102 attached on a KUKA KR 200 robot.

CNC ROBOTIC ORBITAL DRILLING

In the paper presented by Kihlman, et al. at the SAE

2002 Aerofast conference [5], the first version of the
201 PODU, shown in figure 1, was attached on an
articulated robot, the IRB4400, 60 kg payload robot.
That paper presented the first paper on Robotic Orbital
Drilling. In addition to that smaller demo and preliminary
study on Robotic Orbital Drilling, a larger demonstrator
was built at Sowerby Research Center at BAE
SYSTEMS in Filton in the UK in February 2004.
Purpose of demonstrator
The demonstration was held in answer to the call within
the ADFAST project for a practical demonstration of the
CNC Orbital drilling unit mounted into a serial robot cell.
The object of building the cell, apart from demonstrated
testimony of the work and results achieved throughout
the project, was to investigate the possibility of drilling
holes to within aerospace standards with the CNC
Orbital drilling unit mounted on a serial robot, see figure
15. It was understood that both time and financial
limitations within the project would limit the amount of
preparation and testing opportunities available, before
the cell could be built and run. With this in mind a
concerted effort was made by the ADFAST partners to
prepare the cell components to the best of their abilities,
before committing to the demonstration. In practice this
meant building a CNC Orbital drilling unit and testing it
at Novators facilities in Sweden, designing and building
the drilling jig and preparing representative samples of
aerospace CFC and Aluminum at Airbus in the UK, and
finally installing and commissioning a serial robot at the
BAE systems facilities in the UK.
Figure 15: The KUKA 200 with the CNC-ODU used in
the demonstrator
Mass-balancing of the CNC-ODU
One of the drilling Workpackage objectives within the
ADFAST project, apart from actually building the CNC
Orbital drilling unit, was to insure that the unit would be
mass-balanced in operation. This was a concern,
considering that the unit by nature would need to offset
a considerable portion of its rotational mass, to achieve
the dynamic diameter offset function; at the heart of
the orbital drilling concept. Although the unit was mass-
balanced (calculated) on paper, it remained to see if this
concept could be coerced into emitting less than 10N
total, over the entirety of its 0-10 mm offset, 0-40000
rpm spindle speed and 0-500 rpm orbital speed
spectrum. In an effort to examine and fulfill this
requirement, the CNC Orbital drilling unit was mounted
into a bench top test rig, comprising of a force
dynamometer platform for measurement, and axial
suppressors used to mimic the supposed rigidity of the
serial robot cell. The forces emanating from the rotating
CNC Orbital drilling unit were subsequently cancelled by
removing and applying mass where needed. In
conclusion this effort was indeed successful. The efforts
to reduce the forces further were abandoned, only when
it became almost impossible to filter the negligible latent
background vibrations (noise) caused by belt trains,
motor humming and peripheral mechanisms; from the
rotating masses. It is important to keep in mind the
serial robots payload of 200kg and the 150kg total
weight of the CNC Orbital drilling end-effector, vacuum
foot and mounting adapter plate. With this in mind the
assembled cell was required to drill holes within
tolerance in representative samples of aerospace CFC
and Aluminum. Furthermore the cell needed to do this,
unsupported or stabilized by traditional pressure-foot
assemblies or guiding structures. Indeed the only
contact between the cell and the workpiece, apart from
the cutting tool was a simple vacuum shroud exerting no
more than 10-20 N.
Pressure-foot
The CNC Orbital drilling unit produces 10 times lower
cutting forces than Conventional drilling. One issue with
Orbital drilling, however, is the radial forces. The orbital
movement is causing small radial movements. Tests
show that even small radial forces cause the robot end-
effector to oscillate. Similar to the smaller demo in
Linkping with the PODU, a pressure-foot for the CNC-
ODU was developed, see figure 16. The pressure-foot
was simply a plastic piece that was pressed against the
surface of the workpiece to be drilled. At first the
metallic ring, surrounding the cutter, was used as
pressure-foot, but a slight jamming effect in the
mechanism hindered the feeding mechanism to work
smoothly. Instead, the solution with the plastic block was
used. If more time were available, the jamming effect
had been eliminated and the metallic ring would have
been used. The metallic ring is a way to make the tip of
the end-effector slimmer. In this Demonstrator however,
that was not an issue, there was open access to the
specimen. An adjustable spring adjusted the pre-
pressure force.
Figure 16: The CNC Robotic Drilling end-effector
with pressure foot
RESULTS
The test specimen was a 13 mm thick aluminum piece,
see figure 17. Hole diameter was inch. Several
hundred holes where drilled. The end-effector went
through twice, first a rough cut followed by a fine cut. A
rough cut was performed with a feed-rate of 600
mm/min, orbital speed 400 rpm and a spindle speed of
32670 rpm. A fine cut was performed with a feed rate of
400 mm/min, orbital speed 400 rpm, spindle speed
32670 rpm and an increase of eccentricity of 0,15 mm.
For the experiments a KUKA KR 200 was used, see
figure 17.
Figure 17: Test speciment for the CNC-ODU IRB
drilling trials
In Appendix A and B measurements using a Mitutoyo
digital probe is presented. The diameter measurements
were recorded at the largest and smallest diameters of
each hole to reveal ovality etc. Additionally the
measurements were taken in the middle of each layer in
a stack. Appendix A is showing each measured value
and graphical plots of these values are shown in
appendix B. In Appendix C measurements using a
Mitutoyo surface test unit for surface finish
measurements is presented. The measurements in
Appendix A, B and C were undertaken by Novator.
The results from the measured data gave H8 holes
(0,027 mm), hole roundness of 0,013 and a surface
finish bellow 0,68 Ra m, which is well bellow
requirements on for example Lockbolts of Ra 1,6 m.
The end-effector unit weighed 150 kg, where the CNC-
ODU was 105 kg. In
One conclusion is that a pressure-foot still has to be
used to avoid movement caused by the oscillating
cutting forces and, through mass-balancing, only small
centrifugal forces. However, the pre-pressure required
for Orbital drilling is much lower than for Conventional
drilling, hence normality problem in industrial robots can
be avoided.
Our experience and conclusion from this paper
indicates that articulated robots has the following
advantages for drilling:
Articulated Robots are standardized machines
Being mass-produced for car industry makes
them cheap in investment
Drill-jigs are replaced by robot programming
 Robots can be re-programmed, which makes
production flexible
Drilling process planning can be done offline,
saving lead-time in production
An Industrial Robot can be used for other tasks
such as automated fastening, reconfigurable
tooling [7].
Although these advantages Articulated Robots are not
used today for drilling on a large scale in Aircraft
Automation. This paper has shown technology to
compensate for stiffness-problems in industrial robots
by drilling with Orbital drilling, which produces low
cutting forces. Low cutting force is one key enabler to
ensure drilling of high accuracy holes.
CONCLUSION
In this paper Orbital drilling implementation and
evaluation of PODU TwinSpinTM 201 at SAAB Ericsson
Space has been described. Also a description of the
implementation and evaluation of the CNC-ODU on a
standard KUKA IRB has been presented.
A global conclusion of what has been presented in this
paper about Orbital drilling is:
Orbital drilling is a method capable of producing
holes with good hole quality in all the
applications tested
Orbital drilling can give the holes higher surface
quality and cleanliness compared
Conventional Drilling
The waste of parts and assemblies due to
lacking good hole quality is minimized when
using Orbital drilling
Orbital drilling is more difficult to control using
today known handheld equipment, i.e. PODU
TM
TwinSpin 201, because the adjustment and
calibration of the machine is sometimes
circumstantial and expensive because the use
of expensive drill jigs, which should be avoided
in the future.
ACKNOWLEDGMENTS
This work has been a part of the partly funded EU project
ADFAST (Automation for Drilling, Fastening, Assembly,
Systems Integration, and Tooling) and we want to thank all
the people in ADFAST who have supported material to this
paper. Special thanks go to project manager Mr. John
Anderson. Another special thanks go to Novator AB
European Aerospace Manager Mr. Mark Ennis and to SES
production manager Mr. Rolf Thrnkvist. We would also like
to give a special thanks to Mr. Tom Harris at Airbus UK.
REFERENCES
1. Latger, F., Harris, T. and Bjrklund S., Drilling Cost
Model, SAE Aerospace Automated Fastening
Conference & Exposition, October. 1-3, 2002.
2. Harris, T., Johansson, S. and Ossbahr, G., A Study
of the Influence of Drilling Method and Hole quality
on Static Strength and Fatigue Life of Carbon Fiber
Reinforced Plastic Aircraft Material, SAE
Aerospace Automated Fastening Conference &
Exposition, October. 1-3, 2002.
3. Lindqvist, R., Eriksson I. and Wolf, M. Orbital
Drilling of Sandwich constructions for Space
applications, Proceedings of the 2001 Aerospace
Congress (ACE), September10-14, 2001.
4. Helin, P., Jerregrd, H., Robertson, A., and Snell, J.,
"Technologies that make a robot reach Absolut
Accuracy ", Proceedings of the 33rd International
Symposium on Robotics, October 7-11, 2002.
5. Kihlman, H., Eriksson, I. and Ennis, M., "Robotic
Orbital Drilling of Structures for Aerospace
Applications ", SAE Aerospace Automated
to
Fastening Conference & Exposition, October. 1-3,
2002.
6. DeVlieg, R., Sitton, K., Feikert, E. and Inman, J.,
ONCE (ONe.sided Cell End effector) Robotic
Drilling System, SAE Aerospace Automated
Fastening Conference & Exposition, October 1-3,
2002
7. Kihlman, H., Ossbahr, G., Engstrm, M., Anderson,
J. Low-cost Automation for Aircraft Assembly, SAE
Aerospace Automated Fastening Conference &
Exposition, Sept. 20-23, 2004
CONTACT

For further information on the PODU implementation

Orbital drilling is one enabling technology for
and evaluation, please contact Richard Lindqvist.
leight-weight robotic drilling due to low cutting
forces
Richard Lindqvist
SAAB Aerostructures
Orbital drilling has shown promising results in the
Assembly Technologies
comparison to Conventional drilling, especially for
E-mail: richard.lindqvist@saab.se
application such as presented in this paper, but the
process demands more expertise to be set-up
accurately than Conventional drilling.
For further information on the CNC-ODU HA: High Accuracy
implementation and evaluation on a KUKA KR 200,
please contact Henrik Kihlman. PODU: Portable Orbital Drilling Unit

Henrik Kihlman CFRP: Carbon Fiber Reinforced Plastics

Department of Mechanical Engineering
Linkping University
E-mail: Henrik.Kihlman@ikp.liu.se
DEFINITIONS, ACRONYMS, ABBREVIATIONS
CNC-ODU: Computer Numerical Controlled Orbital
Drilling Unit
PODU: Portable Orbital Drilling Unit
BB: Bread Board
QM: Qualification model
APPENDIX
A drilling report from drilling trials using a KUKA KR200
IRB as manipulator and a CNC-ODU Twinspintm 102 as
a drillunit, Compiled 2003-11-25 by Novator AB.

SES: SAAB Ericsson Space

APPENDIX A
APPENDIX B

Figure B.1

Figure B.2
Figure B.3

Figure B.4
APPENDIX C

Figure C.1