Professional Documents
Culture Documents
Henrik Kihlman
Production Systems
Department of Mechanical Engineering
Linkpings Universitet
Linkping, Sweden, 2005
ISBN: 91-85299-59-6
ISSN: 0345-7524
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 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 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
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
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).
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.
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.
7. System Setup Explains when calibration is required; and presents the calibration
Calibration calculation procedures.
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.
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
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.
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.
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
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.
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
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
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:
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.
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
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:
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:
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.
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
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.
16
Introduction
NFFP2+ FlexAA
ADFAST FINABOX
CRC- IMST
Academic Relations
- PROPER
- ProViking
- Conferences
- Licentiate thesis
- Collaborative research
Student Project #1
Student Project #...
Student Project #27
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:
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 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
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:
The tooling, drilling and metrology suppliers, in the ADFAST project, prepared the output of
the project for exploitation. More specifically their roles were:
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.
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
Fuselage
Horizontal
stabilizer
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
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.
25
Aircraft Production
Sheet metal
Order Part assembly System
planning integration
Composite
manufacturing
Final painting
Delivery to
customer
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
Composites 10%
Aluminum 75%
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.
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
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:
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:
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.
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
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.
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.
B - Work envelope
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).
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.
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.
2.6.1 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).
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.
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).
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 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.
2.6.4 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.
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:
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):
43
Aircraft Production
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
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.
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.
48
Robotic Orbital Drilling
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.
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).
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
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.
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).
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.
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.
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.
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
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
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
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.
55
Robotic Orbital Drilling
Cutting edges on
both side and face
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 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
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.
57
Robotic Orbital Drilling
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
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
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
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
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
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.
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.
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
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
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.
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).
65
Affordable Reconfigurable Tooling
Workpiece
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.
67
Affordable Reconfigurable 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.
68
Affordable Reconfigurable Tooling
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.
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
Dynamic
Modular Modules
Framework
Metrology Box-joints
system
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.
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.
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.
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.
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.
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.
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 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.
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
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.
77
Affordable Reconfigurable Tooling
Male Capto
Female Capto Chuck
Chuck
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.
78
Affordable Reconfigurable Tooling
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.
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
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.
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
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.
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
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.
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.
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
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.
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.
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
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
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
Beam displaced
Coaxial state
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.
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
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.
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
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
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
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
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.
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
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
97
Metrology for Maintained Accuracy
Orientation in space
from LED images
Position in space by
reflector tracking
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.
98
Metrology for Maintained Accuracy
2 out of the 10
Female Capto LEDs
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
99
Metrology for Maintained Accuracy
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.
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
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
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.
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
106
System Integration
All the functions in the generic robot program include the following events on NextAction:
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.
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.
ObjConnect.ConnectEmbeddedSystem 192.168.0.1
ObjAsync.SetContiniousTimeModeParams LaserInterv
ObjAsync.SetMeasurementMode / 1
ES_MM_6DContiniousTime 2 LTConnect
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:
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
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.
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
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
Brown: Calculated
Red: Coordinates
Orange: Laser Calibrated
{6DPrism}
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.
113
System Integration
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
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
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.
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.
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.
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
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
q0Act q0Nom
q1Act q1Nom
Err = 2 arccos .
q2Act q2Nom
q3Act q3Nom
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
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.
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}
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.
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
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.
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.
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.
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
Tr Tr Rob TCP 0
6 D Pr ism = Rob T TCP 0 T.
6 D Pr ism
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.
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
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
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
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.
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
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:
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.
129
System Setup Calibration
Tr
DM = Tr6DPrismProbed 6 DFemale
T
6 D Pr ism
-1
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
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
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:
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.
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
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:
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.
Holes_dia() =import(wing_section_A_holes)
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:
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:
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.
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
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.
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
Step 3
Analyse Gravity & Process Forces
Step 4
Robot Selection & Reach Analysis
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
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
144
Operation Planning Methodologies
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.
145
Operation Planning Methodologies
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
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.
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:
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:
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.
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.
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.
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.
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.
10.3.1 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.
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.
159
Future Improvements
of the system will be affected. Some parts in the digital model of the system are less important
than others.
160
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.
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.
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.
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
// & & / /< & - &
?& ) ! ' (
9 .4 0 .@ '? "1 A,&
( ' & 0 = - 6B # + (
4 -7
// & & &
: : & *' + ,+ 45
.4 0 .@ '? '#8 "C /( " D 6
169
References
<= 5 -5 8 - * B -< + * ( )
4 -7
// & <& / 5
/ < I / $ !/ &4 5&
<3 = 9 : % 5 : ( +55 5 5 * *
& <& / 5/ &
5 F
G 8& & ,2, " " ' # , (( 2* A&
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 "&
170
References
= 4 &D& * = + < + 7= 4
& @ : @ +55 D 69 5 5&
" 6 5- % 5 * % 5 6 + -5 *
4 -7
// 4& & / / 5Q ' R, * )*2>S 6R 5<6> R &
" : @ 5
5 8& 55< 4' 5
9 ; 5.4 0 .@ '? "1 ,)2
; 9& * '
.4 H + 6 B 5
5 < <&
171
References
8 <4 &"& B4 45
5 &8& ,A 7 ! ,4 ' & . 4& = <&
;&'& &'&.& & -& !&
D4 8&B & B 5
: @&;& ,,2 @
A ? , ;51 5 ' 5
5<
9 <= 5 H + 6MD 5
6 U) 5 % < H &
172
References
? 45 @& , ! '
' !0 .4 # & ,2 0 H ."? 0'= 7*) -
9 4 5" < < 0 :G- < H + " *2 2) 0 :G- < ' #,
!)!) (A 2
9 4 & ) ' * 7
= - > =4 < 5
&! # ) D - -- &
9 4 & 4 & 4 7
/ ' 7 0 ' = H .
= % 5 - # & A&
173
References
# 6 D& * = + < + 7# 6
D& @ : @ +5
5 D 69 5 5&
# 4 5B & 7 0) " -
; <= > "1-
5
5 : @& ,!A ' # 2!, , 2( A
4 & 9 4 / '
7<< ! 4 - 8 < -46 H + 6
5 5 #V ( A H &
174
References
5
5 & # +< & ) 1 '' 1 0
' #7 ) 2 )
&
) 0 : 5 5
69
; 5.4 0 .@ ' V "C )(! ) 0 :G- < +
B4 "& #6 9& 8
" - ; <= > "1- &
175
References
B 5 (? % 1 2 " 2> 2
! ' ## &, 2( * H # &
176
Appendices
Appendices
Contents Page
A: Adjusting a Hexapod to High Accuracy 178
177
Appendix A: Adjusting a Hexapod to High Accuracy
Iteration
stops
Iteration
stops
178
Appendix B: The Hydraulic Diagram
179
Appendix C: A Short Summary of Transform Equations
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:
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 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:
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
Cone
Figure D.1: The Force Cone developed by the author and University of Wollongong,
Australia
182
Appended Publications
183
Appended Publication I
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.
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.
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.
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
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.
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
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)
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).
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.
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 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.
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
Also the aircraft assemblies are becoming less Since tool reconfiguration times not are critical in
complex in terms of number of parts and geometric aircraft assembly it is fully possible to let an external
variation thanks to modern high-speed machining. device reconfigure each dynamic module in turn.
There will in other words be less but larger parts to The new concept is based on the use of a standard
assemble which will make it easier to obtain enough industrial robot to sequentially reconfigure each
free space in the tools for robots and other equipment. dynamic fixture module. The robot is cost-effective
In other words it is now better opportunities than since the idea is to also use it for assembly process
before to succeed with reconfigurable aircraft operations such as drilling, countersinking, riveting
assembly tooling. and materials handling. The dynamic modules of the
tool in turn are attached to a static framework and
2.2 Possible tooling concepts for airframe they have built-in locking mechanisms to hold the
assembly configuration between the set-ups. A principal sketch
Flexible tooling in general has been used for a long is shown in figure 1.
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 Figure 1: A principal sketch on the ART concept
based on the ability to re-configure the tool in just a showing the static framework and the dynamic
few seconds so that different assemblies within for modules which attach to aircraft parts.
instance a product family can be assembled or
machined. Normally this type of tooling is only used
for machining operations and not for assembly. The static framework represents a universal tooling
In 1990, Goodman (13) mentioned that future aircraft system together with the dynamic modules. Each
assembly tools probably would be reconfigured by dynamic module can not only be re-configured with 6
the use of servos and encoders to perform the set-up degrees of freedom but also be moved to another
change of the fixture. Similar ideas were also position on the static framework. This latter process
presented by Li et al. (7) in 1996. is here called rebuilding. Rebuilding will only be
performed if the current configuration of the tool
Today there are several suppliers of this kind of cannot handle the geometry of the next product.
"hyper flexible" tool systems. One manufacturer of Assembly analysis and programming will be done on
this type of tooling is MTorres in Spain (11). Another an offline simulation and programming system
type has recently been presented by Kostyrka and (OLP). Such systems are normally used for the
Kowalsky (6). They have also taken part in programming of the robot but the kinematics of the
discussions of the use of passive dynamic modules, dynamic modules can be defined and simulated as
meaning that the movement in the dynamic well. Figure 2 shows a model of a tooling with its key
(movable) module would have to come from an components,
external source, such as e.g. a robotic device. The
dynamic module only withheld the configuration 3.1 The re-configuration process
between the set-ups through a built in locking
mechanism. This principal idea has also been studied A simplified description of a reconfiguration process
in an EU-funded project called "Flexible Workshop is as follows:
for Airframe Assembly" described by Engstrm (2). 1. Using the OLP, position the workpiece(s) inside
the static framework without any dynamic
modules.
3 A NEW RECONFIGURABLE
TOOLING CONCEPT USING AN 2. Study the position and direction of all pick-up
points. Re-build if necessary.
INDUSTIRAL ROBOT
3. Locate the appropriate positions of the dynamic
By joining some ideas from modular tooling and modules on the static framework to get access to
hyper-flexible tooling based on the idea of letting an each pick-up point.
external device position passive dynamic modules, a 4. Make a FEM analysis of the tool configuration.
new reconfigurable tooling concept is being
developed. This tooling concept is called an 5. Simulate and program the robot trajectories
"affordable reconfigurable tooling" (ART) and the between the previous and the current
very first ideas have been presented by Kihlman (5) configuration of the dynamic modules.
and by Kihlman and Engstrom (8). 6. Simulate the drilling and the riveting operations
for each stage
Proceedings of the 33rd ISR (International Symposium on Robotics) October 7 11, 2002
or very few tools instead of one conventional tool per product specific parts need to be used such as e.g.
product variant. The economy of a reconfigurable contour boards.
tool here lies in that it can be used for similar parts In this economical comparison the costs for
of different products. conventional tooling have been estimated by
Small scale manufacturing gathering economical numbers from the real
development of a tool at SAAB Aerospace.
Some aircraft are produced in very small volumes
compared to for instance short-distance commercial The estimation of specific and common costs for
aircraft. This could be the case when production of reconfigurable tooling has been made by configuring
special military aircraft, certain commercial aircraft a tool concept for the same assembly as above and
models, flying demonstrators or even prototypes are trying to estimate the costs for the parts and the
concerned. In those cases the total manufacturing design and the manufacturing of it. Since the concept
volumes could be between 1 to 100 units. In such is based on the use of an industrial robot and an
cases reconfigurable tooling can be very economical advanced measuring system it has also been
if it is possible to use it for different airframe parts necessary to handle these costs. This was solved by
of the same product. simply estimating how much of the investment costs
of these that could be allocated to the reconfigurable
Manufacturing ramp-up tool. Our estimation was that less than 20 % of the
A third interesting application is when the peacetime robot and measuring system costs could be allocated
production rate of a military aircraft is very low and to the tool. In other words 80 % of the robot cost
then suddenly a very high production rate is wanted. should be allocated to assembly process costs.
In this application a reconfigurable tooling can be The result of the estimation of costs is shown in
both economical and a significant enhancement for a figure 3.
fast and dramatic increase in the production rate.
As seen in the figure a reconfigurable tool for the
A reconfigurable tooling will also make it easier to actual structure will need almost three times the
continuously update the product since design investment of a conventional tool. Of this investment
modifications are easier to perform. a small amount is product specific costs that will
occur for every new product.
4.2 Risks and weak points
Even if reconfigurable tooling is likely to have Investment costs for re-config. and conv.
economical advantages there are also some weak tooling
points and risks that have to be considered during the
800
continuous development. A very critical technical
700
issue, stiffness, has already been pointed out. 600
Money unit
same assembly content and each tool has the a rate of 20 per year is needed. If it were possible to
maximum capacity of 200 units per year. A tentative produce five major airframe assemblies in the same
production scenario in terms of products and volume reconfigurable tool this would mean that five
over 15 years is illustrated in figure 4. 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
Produced units per product and year can be used also for other products. There is a lot of
500 I free capacity (100 units/year) as long as only 20 units
H per year of these five assemblies are produced.
400
G The economical analysis becomes more complex if a
Volume
11
13
15
scenarios reconfigurable tooling will definitely be a (10). Kihlman, H., Eriksson, I., Ennis,. M.,
very strong economical alternative with some very Robotic Orbital Drilling of Aeronautical Structures
interesting properties. and Materials, SAE Aerospace Automated Fastening
However, there are still many technical aspects to Conference & Exposition, 2002
analyse e.g. stiffness and the overall system (11). http://www.mtorres.com/ingles/ia/index
functionality (tool, robot and tracking system) has to .asp
be tested in reality. Therefore, prototypes of
reconfigurable tools are currently being manufactured (12). Lindqvist, R., Ericsson, I. and Wolf M.,
for evaluation. This and other further work will "Orbital drilling of sandwich constructions for space
continue over the coming years. applications", , SAE Aerospace Automated
Fastening Conference & Exposition, 2001
In the end, this work hopefully will lead to an
increasing automation level in aircraft industry with (13). Goodman, H., "Assembly cell speeds
more industrial robots. And this at the same time as riveting", Journal of tooling and production, Dec.,
manufacturing flexibility is increased in order to meet 1990
the future production scenarios of aircraft industry.
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
Magnus Engstrm
Commercial Programs
Saab AB
The paper starts out with a brief survey of the The conventional way of building assembly tooling in
conventional way of building aircraft assembly tools aircraft assembly is to use steel material, which is rigid
followed by discussing existing solutions of tooling that and has good thermal expansion properties. Steel is
has flexibility. Finally the paper presents the general also cheap and has good welding characteristics.
idea of the ART-concept and the modules included. Aluminum is also utilized in conventional tooling,
particularly for movable fixture parts and turning fixtures
CONVENTIONAL TOOLING where low weight is preferable. In a typical development
of a conventional tool, the tool is designed with input
Production tooling in aircraft assembly is divided in two from the geometrical data of an aircraft structure that is
categories. The first one is jigs. Jigs are workholders to be assembled. Although this data is needed in order
designed to hold, locate and support a workpiece while to understand what the fixture must look like, the
guiding the cutting tool throughout the cutting cycle [2]. designer must start the designing of the tool before this
The other one is fixtures, who are workholders which are geometrical data is available. This must be done in order
designed to hold, locate and support workpieces during to shorten the lead-time for the tool parts, which
the machining cycle [2], in this case holding the generally is long. The parts with the longest lead times
workpieces during assembly i.e. drilling followed by must be sent to the manufacturer before the tool
riveting. Wang [3], defined fixtures as consisting of designer gets the final geometrical data from the aircraft
clamps and locators, that must be capable of designer. This is one of the reasons why there are a lot
positioning, holding and supporting a workpiece during of changes involved in the development of conventional
machining, assembly and inspection aircraft assembly tools.
Boring and drill jigs are something that aircraft producers To stay as a competitive aircraft manufacturer in the
want to eliminate nowadays, and instead focusing on future the tools must be standardized and flexible. The
automation using robotics or using CNC predrilled holes. assembly system must also be able to handle different
Sometimes it is difficult to tell the difference between jigs product types within the same system. This can only be
and fixtures. Conventional Tooling is a comprehensive done if the flexibility and modularity is increased
term for how aircraft assembly fixtures has been, and throughout the assembly system.
still are designed.
FLEXIBILITY AS A SOLUTION
Conventional tooling is characterized by that they are
tailor made for each assembly. If a new product is Flexibility in the production area and flexible production
developed a completely new conventional tool must be systems has been spoken of since the beginning of the
designed. This is why they sometimes are called 1980. One clear reason for that is that most
dedicated tooling [4] or hard tooling [5]. Conventional manufacturing industries more or less have experienced
tooling however is well known and they have been rising number of product variants and decreased product
practiced since aircraft started to be built in aluminum. lifespan. Nowadays, more different units need to be
constructed in smaller quantities, which increase the
Pins and clamps are used to locate and hold airframe demands on the production systems. The system is
parts in position before these are drilled and fastened supposed to manage more product variants and the
into one unit [6]. The pins and clamps are often ability to change product generation frequently.
positioned on pick-ups. The pick-ups are elements
between the aircraft structure and the frame of the jig. In In the Visionary Manufacturing Challenges for 2020 [7],
extreme, the design of the pick-ups is generally unique moving from the current status of manufacturing to
for each assembly. For the pick-ups there are rarely any manufacturing in 2020 there are six grand challenges
standards for the designer to use. that would make the vision possible. One of them is to
make the manufacturing enterprises reconfigurable to
Designing new elements for each new assembly is rapidly response to changing needs and opportunities.
expensive and time consuming. Although conventional This implies that the manufacturing systems must be
tooling is safe and well known and almost anything can adaptable and reconfigurable and have a flexible
be done due to its tailored design, the time and money forming process so that no hard tooling is needed.
needed are two elements the aircraft companies do not
have a lot of nowadays, as the competition between The problem with companies using product- or variant
aircraft industries has increased. Earlier, aircraft specific tools is that the cost needs to be distributed
manufacturers made their own complete aircraft, and among fewer products, which gives heavily rising
now they are becoming specialist in a specific field of the product costs as a result. A report from Tichem et al [8],
airframe. One example of a field is wing structures. defines flexibility as the capability to react to changes in
Although computer enabling digital mockups and the environment. The same report discusses the
advanced machines with high speed machining problem with automation in the assembly process; it
capabilities has become general practice, the often leads to the development of case-specific
solutions. It results in high degree of assembly
engineering and high capital investments for each Further in this paper three levels of flexibility is going to
product to be assembled because of product-specific be presented. The first level of flexibility represents the
equipment that has to be developed and used. This ability to adjust a fixture for minor changes within one
phenomenon does not permit the application of product. The second level of flexibility represents the
assembly automation for small series production. ability to reconfigure the fixture between parts of the
same product family. The third level of flexibility is the
Flexible fixturing involves employing a single fixturing ability to be able to rebuild the fixture between product
system to hold work-pieces of various shapes and sizes families.
[9]. A new paradigm was presented by Mehrabi et al.
[10], called reconfigurable manufacturing system (RMS), CURRENT SOLUTIONS
which they defined as the next step after Flexible
Manufacturing Systems (FMS). They stated that RMS is Flexible Fixturing is not new. There exist several ideas to
designed for rapid adjustment of production capacity and increase the amount of flexibility. The first approach is
functionality, in response to new circumstances, by Modular Tooling. Flexibility in this paper is defined as
rearrangement or change of its components. Applying how quick the system can be reconfigured on a macro
RMS will go beyond mass, lean and flexible scale. Macro scale is the ability to build vehicles as cars,
manufacturing towards reduction of lead-time for aircrafts etc.
launching new manufacturing systems and also enable
rapid upgrading and quick integration of new process MODULAR TOOLING
technologies and functionalities into existing systems.
Modular Tooling represents the first level of flexibility in
A product family is defined as one or more part types assembly tooling. The idea is based on building the tool
with similar dimension, geometric features, and by a standardized kit of parts and couplings. Today there
tolerances, such that they can be produced on the exist several suppliers for Modular Tooling who have a
same, or similar production equipment [10],[5]. In figure wide range of parts and can ensure a certain quality.
1, the three most common product families in aircraft Modular Tooling has been used in the car industry for a
production are presented. long time and is recently being introduced in aircraft
industry.
Modular Tooling is a good example of a technology Increasing the level of flexibility to the third level of
where the system can handle changes within the same flexibility move us to CNC-controlled tooling or hyper-
product type, i.e. the first level of flexibility. The fixture flexible tooling. CNC-controlled tooling has the ability to
can change in the range of centimeters; therefore it is configure the fixture for different structures similar as
not possible to do any reconfiguration to assemble Flexible Tooling, however all moveable fixture units can
different types of structures with one tool configuration. be configure simultaneously. Goodman [5], suggested
However, the tool can be dismantled and rebuilt to this as the next step for flexible assembly, where the
assemble another type of structure. But this is rather an flexible jigs position themselves using multiple-stepper
example of reusability rather than an example of motors. One complete reconfiguration of the tool is
reconfigurability. Rebuilding the tool would take a couple possible to do within a couple of minutes.
of weeks.
A similar project of this research was done in the FAS
FLEXIBLE TOOLING (Flexible Assembly Subsystem) program presented by
Wright and Sarh, [19], and Olsen [4]. They had the
Flexible tooling is the first step towards increased scope to increasing quality while reducing costs. They
flexibility. Flexible tooling is able to reconfigure the developed and demonstrated flexible and automated
movable fixture units sequentially one by one until the workcells that could replace dedicated tooling and be
entire tool is reconfigured. The can be reconfigured for a able to quickly reconfigure itself for new types of
new product within the same product family. subassemblies. A robot was used for assembling the
Reconfiguration of all fixture units is possible to do within aircraft parts and the flexible fixture consisted of four
the hour. three armed robots with a total of 48 axes controlled by
the fixture controller. Where each of the 12 arms has a
A lot of research was done in flexible tooling back in mid ball-joint mounted between the last link and the end-
80th and early 90th. One of them was at McDonnell effector quick-change unit. The fixture controller can
Douglas, Long Beach, CA, where they developed an either release or freeze the ball-joint unit. Its end
automated flexible assembly cell to gain flexibility and effectors are either clamps or suction cups mounted on
reduce tooling costs [5]. They replaced the separate quick changers.
tools with a single, flexible tool that serve a family of
assemblies and provided automated inspection. The cell A more recent approach to CNC-controlled tooling was
automatically assembles, rivets and inspects small, flat presented by Yingjie et al. [20]. Yingjie had an approach
bulkheads. They used a gantry robot to configure the on a flexible workcell for aircraft wing spar assembly.
flexible jig and loaded detail parts into the jig. Goodman They developed a concept of a fixture with 76 step
claimed to replace 250 conventional jigs with one flexible motors controlled by a central controller. The concept
tool. was suitable both for different type of aircrafts. and also
for the left wing and the right wing of the same aircraft.
Manufacturing of sheet-metal parts is considered difficult
to automate because of the complex geometries This work with CNC-controlled tooling has been carried
involved. This problem was solved by M.I.T. in mid 80th on by Kostyrka, [21], and MTORRES [22]. They use
by Youcef-Toumi et al. [12],[13], Youcef-Toumi and pogo sticks, which consists of a cylinder, which can be
Buitrago, [14], Fields et al. [15], Asada and Fields, [15], extracted or reversed in one axis. A clamping device is
and Asada and By, [16]. They used a method called used to lock the cylinder. The sticks also have servo
Robot-Operated Adaptable Fixture to change the engines and encoders to locate themselves to the right
configuration of a fixture bed by manipulating vertical configuration. The sticks are put together into a matrix,
supports (fixture elements). Changing the length and forming a bed of nails, se figure 3.
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.
A NEW TOOLING CONCEPT
DYNAMIC MODULES
MANIPULATOR INTERFACE
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
2002-01-2636
Many different ways have been tried to compensate for ORBITAL DRILLING
dislocations in robots today [9],[10], i.e. where an
external measuring loop is used to improve the final Orbital drilling is based on moving the tool
position of the robot TCP (Tool Center Point). These simultaneously in both axial and radial directions by
methods are basically about moving the robot to a feeding the tool through the material and causing the
position in space and when the position is reached the tool to rotate about its own axis while at the same time
measurement system can verify and correct the position. revolving the tool. Since the center of the tool orbits
Gooch [9], stated that metrology is an enabling about a stationary hole center the thrust force is
technology and applications are being extended from minimized and many of the problems associated with
inspection to control of the manufacturing process itself. traditional drilling (stationary tool center) are eliminated.
A state-of-the art laser measures down to 50m.
However, a state-of-the art IRB has an incremental Moreover the tool diameter is less than the hole
accuracy of 70m, which limits the positional accuracy. diameter enabling efficient chip and heat extraction. The
cutting edge is partially and intermittently in contact with
Even if a high enough positional accuracy could be the work piece allowing for efficient chip extraction. The
achived in the robot; the ability to adjust the robot to this eccentricity is adjustable, thus the precision of the hole
is limited in time. Every time a correction of positional is independent of the precision of tool. This in turn
accuracy is carried out the measuring equipment sends makes it possible to drill high precision holes with low
data to the robot. This is called synchronization. Hence, precision tools and to drill holes of different diameters
even if a laser can measure the position of the robot with tool of one diameter. As seen in Figure 1, orbital
TCP in 1000 Hz, this does not mean that the robot can drilling can machine different kinds of axis-symmetrical
be corrected at that speed. This is due to the limitations holes such as cylindrical, conical and complex shaped.
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
EXPERIMENTS AND RESULTS
FORCE MEASUREMENTS
orb_05
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 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
Orb_d_01
Sample (10.000Hz)
Vacuum boost Z
attachment
Y
Orb_d_05
Sample (10.000Hz)
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
2003-01-2961
Raimund Loser
Leica Geosystems AG
WEBWARE
POINT-TO-POINT CONTROL
x Alt 2.
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
look at the robot used for the experiments - an singularities using the pointing mode. When joint 4
IRB4400 from ABB. The repetitive accuracy in that and joint 6 is passing through zero axis degrees, these
robot is 50m. To reach that accuracy in the full work kinds of robots will cause strange behaviours.
volume, the resolution of the robot is much higher. Pointing mode is also difficult for the robot to
Tests performed in this research have proven the manage if the weight of the end-effector is to high,
robot to able to move in a resolution of 5um. When which will cause a torque on joint 5. For example, the
fine adjustments have been carried out, the robot IRB4400/60 used in the experiments manages 60 kg
controller is simply given a new absolute position in at maximum 27 cm (16.2 Nm) out from the TCP.
the base coordinate system. Helin, 2002 (3) presented
tests results from ABBs High-accuracy program. 2.3 Friction in workobject
The results clearly showed that it is easier to reach Also in this research the robot has been tested as a
higher accuracy in smaller robots. One trend in manipulator for flexible tooling (4),(5). The robot is
Aerospace Industry nowadays is to buy large robots docking with tooling modules and configures them,
for being able to carry multi-end-effectors weighting see figure 2. The tooling modules presented in that
several 100kg. It is contradictory to use large robots research can be manipulated in 6DOF. Using the
when high accuracy is required. This indicates that robot to change the configuration of a mechanism
end-effector suppliers should consider building however is a challenging task, especially in fine
smaller end-effectors if they are to be fine adjusted, adjustment. There was a clear indication that it is
which requires smaller robots and the question arising more difficult to position a robot in high accuracy if it
to the end-effector suppliers today is: How will large is exerted for friction. Tests performed in this
end-effectors be positioned with high accuracy? To research verify this hypothesis. In general, the time to
give better understanding to the background to this reach high accuracy becomes longer if the robot is
research, the next two section will presented why moving a fixture module. Some tests have also
extensive calibration procedures will not help to resulted in overshooting when compensating the
maintain high precision in some scenarios. datum point of the fixture module.
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. Figure 2: Robot-manipulated Tooling
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 The conclusion from this section is that even highly
hanging mode on the robot to avoid unnecessary calibrated robots would not help in applications
torque in joint 5 such as figure 1a. When applying where the process force changes rapidly or if friction,
pre-pressure force on the end-effector joint 5 will be which is not static, is involved. One solution to cope
exerted to high forces, in some cases more force than with these problems is to have a metrology system
the weight of the end-effector itself. 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.
(Absolute Distance Meter), (8),(9). The ADM enables equipped with an automatic hydro-mechanical Capto
the tracker to locate individual target reflectors chuck system. This enables the robot to interact with
without the need to measure from a calibrated origin flexible tooling modules presented by Kihlman and
position. The ADM can independently measure the Engstrm, 2002 (5), ordinary drilling machines and
absolute distance to any unknown reflector location, Orbital Drilling machines such as presented by
which simplifies re-initialising the interferometer. Kihlman et al., 2002 (7). The controller on the robot
Whereas the ADM is only used in initialising the used for the experiments was the S4Cplus version.
interferometer, it cannot itself be used for tracking.
4 COMMUNICATION
The LTD800 can have an additional camera, the T-
In order to connect a metrology system with a robot,
Cam, which is attached on top of the tracker unit. The
the robot needs to be able to communicate with the
T-Cam measures the orientation of a target. The
metrology system. ABB robots, which have been
accuracy of the LTD800 is 10 ppm (m/m) for static used in this research, have today basically two
targets, and 20 ppm for moving targets. Distance communication ports. One is serial port, which has
resolution is 1.26 m and angle accuracy is 0.02 been a standard for communication with robots in
degrees. The angle accuracy is kept constant over the general for many years. Today exists also TCP/IP,
full measurement volume, through the use of a zoom which now has become a standard for communication
objective in The T-Cam. The zoom objective with robots. RS232 and TCP/IP has until now mainly
continuously maximizes the resolution from the been used for downloading robot programs from an
reflector targets in the camera picture. offline programming system. Presented in this paper,
TCP/IP has been used for synchronising a servo loop
3.2 The 6D-Reflector
on an external PC with a metrology system and the
To measure robot positions in 6 DOF a reflector was robot controller.
developed. The 6D-Reflector is a further
development of the T-Probe that is a commercial 4.1 WebWare
product from Leica Geosystems AG. The 6D- The communication with the ABB robot was
Reflector is always initiated with the ADM, since achieved through WebWare. WebWare is a driver
there is no birdbath available for the 6D-Reflector. package from ABB that uses ActiveX controls and
The 6D-Reflector comprises of an aluminium housing OPC (OLE Process Control) server to enable a
that is attached 10 LEDs and a prism. The LEDs network connected computer to manipulate RAPID
enables the T-Cam to measure the orientation of the program execution, RAPID variable data and I/O-
6D-Reflector. signals on the S4 controller.
4.2 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
Male Capto is accessed over the network through conventional
Female Capto
socket communication. A call for measurement is
Figure 3: The Multi-purpose 6D-Reflector initiated, and a response is given. This execution can
either be performed synchronously, where the
The LEDs are positioned at different depths for the program is stopped until a measurement is sent back
camera to measure orientation. The LEDs are flashing to the client, or asynchronously where a trigger is
in infrared light and the camera is zooming in the 6D- activated when the data is sent back to the client. The
Reflector so that the LEDs are taking up the full LTD800, according to specification, is able to
picture. The laser beam is reflected back to the measure 6DOF in 100 Hz and can interpolate
tracker unit from a glass prism retro-reflector. The between measurements in 1000 Hz. To use
glass enables the beam to be +/- 50 degrees away measurements for feedback control however, requires
from the prism axis before contact is lost. The 6D- each measurement to be sent by emScon to the PC
Reflector housing has two ergonomically designed prior to updating the robot. To clarify this, the
handles when moving the 6D-Reflector manually. For emScon server is sending 4 packages with data per
attaching the 6D-Reflector a system called Capto C4 second. This means that the 100 measurements per
from the Company SANDVIK AB was used. The seconds are divided in 4 packages of 25
Capto system is generally used for attaching cutting measurements. For a feedback control loop, only 4
tools in NC-chucks. The repetitive accuracy of the measurements can be used for updating the robot
controller per second, hence the rest is just old data.
Capto C4 is 2 m with a maximum load capacity up
to 17 kN. This indicates that the bottleneck in the integration
now is the emScon server. Updating frequency in the
3.3 The Robot system is hence 4 Hz. For the applications presented
in this paper on the other hand, this update frequency
The robot is a standard Industrial Robot, the IRB4400 was enough to manage the processes. If however the
from ABB. The IRB4400 has a repetitive accuracy of robot needs to be updated at higher speeds, such as in
50 m and a maximal payload of 60 Kg. The robot is
Proceedings of the 35th ISR (International Symposium on Robotics) 23-26 March, 2004
compensating deflection in drilling, this speed would spikes sometimes up to 50 ms. It was clearly
not be enough. 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
The LT-controller plus includes two independent
second, another issue would appear in the integration;
computers. One is emScon that runs on a Windows
Robots have their own controller with ramp-ups and
2000 operative system, which is not a real-time
brake patterns that will slow down the actual step-
system. The tracker processor is running on a real-
response time. For rapid update of the robot
time system and could support a high-speed interface.
manipulator another method would be required, such
It could be solved in two ways; either through a
as through shared memory communication directly
parallel interface using IEEE 1284 that would enable
with the robot controller.
data transfer with a very short delay and low
overhead. Transferring a measured point would with 4.4 Programming method
this method take only 100-200 microseconds.
Another method would be to use a 100Mbit/s or All robot activities were programmed in the Offline
1Gbit/s Ethernet with point-to-point connection, simulation/programming package V5Robotics/R12
using the UDP protocol. Measurements could then be from DELMIA. Since the robot in this research does
transferred within 1-2 milliseconds. Neither of these not operate like normal robots, the post-processor
methods is supported today, but it could be possible phase into RAPID code was not done in the OLP-
to accomplish. Changing communication method to module as normal. Instead the XML-code prior to this
real-time would as a consequence also require phase after all simulations and was downloaded to the
another platform for the Server PC, which today runs Server PC of the metrology-integration and
on a Windows 2000 operative system. Transferring interpreted there. This enabled the programming to be
measurements in real-time would be triggered at a fix performed on a more task-oriented approach. On the
rate based on a clock, or a trigger line, whereby robot existed only one program that was generic and
transferring of data is done with only a short delay. never changed. The Server PC executed different sub-
TCP/IP and emScon would still be used on a real- routines such as docking, drilling, fixture
time platform, but only for things such as setup of the configuration etc. The program execution on the
tracker, starting and initialising tracker and camera, Server PC manipulated, through WebWare, different
establish transformation parameter from the tracker to PERS (global) variables, reading and writing I/Os etc.
the object coordinate system etc. on the robot. WebWare does not support direct
manipulation of move commands. The generic
4.3 Integration program concept enabled this to be overridden.
The integration was implemented with a standard
TCP/IP network set-up. An Ethernet switch connects 5 CALIBRATION METHODS
the PC with the robot controller and the emScon
Before starting the execution of the metrology
server, see figure 4.
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
Figure 4: The system architecture for the integration Chuck
transforms is interpreted as the coordinate { B } in the important to realize that this calibration will not affect
{ A } coordinate system. the final positional accuracy of the robot, but merely
the time to reach a high-accuracy position, since the
5.1 Robot chuck servo loop on the server PC will continue iterate until
The first step in the calibration was to calibrate the a position is reached within the specified tolerance
Re fl
interval.
transformation Chuck T , which is where the chuck TCP
is positioned relative to the 6D-Reflector sitting next
Another method of calibration was also tested which
to the chuck. Calibration of the Capto TCP is most
again moves the robot to different configurations.
sensitive to avoid tolerance build-up, which results in
The kinematics equation to be solved is:
low absolute accuracy. The method is summarised as
Re fl 0 TCP 0 Re fl 1
the following: TCP 0 T TCP
TCP 0 T Re fl 2 T = Tr T Re fl 2 T
Tr
1. XY-plane: Measure the plane of the pressure- to the project manager of the ADFAST project, John
foot using a reflector ball Andersson. We would also like to thank SANDVIK
2. Z-direction: Measure a cylindrical object in the AB for showing great interest and providing to us the
same direction as the drill on the machine. In Capto System.
this case it is not important that the cylindrical
center is align with the drill Z-axis. 8 REFERENCES
3. Z-center: Position the reflector ball in the drill
bushing. The reflector ball does not need to be (1). DeVlieg R., Sitton K., Feikert E. and Inman J.,
on the XY-plane. The measured point is simply ONCE (ONe.sided Cell End effector) Robotic Drilling
transformed to the XY-plane. System, SAE Aerospace Automated Fastening
Conference & Exposition, October 1-3, 2002
The transform cannot be measured directly but
Tr
T and Tr
T can be measured. Then Re fl
T (2). Craig J. J., Introduction to Robotics Mechanics
Bushing Re fl Bushing
and Control, Addison-Wesley, ISBN-0-201-09528-9,
is calculated according to: 1989
Re fl
Bushing T= ( Tr
T
Re fl )
1
Bushing
Tr
T (3). Helin P., Jerregrd H., Robertson A., and Snell J.,
Technologies that make a robot reach Absolut
Offline-calibration of End-effectors Accuracy, Proceedings of the 33rd International
Probably the most reliable and obvious method to Symposium on Robotics, October 7-11, 2002
calibrate the TCP of an end-effector is to measure the (4). Kihlman H., Affordable Reconfigurable Assembly
unit in a CMM (Coordinate Measurement Machine). Tooling - An Aircraft Development and
A CMM has an absolute accuracy in the order of 10 Manufacturing Perspective, Licentiate Thesis No.
m, hence presumably better than a hand-held 980, LiU-TEK-LIC-2002:53, Department of
reflector ball. Basically the same transforms as the Mechanical Engineering, Linkpings Universitet, SE-
Inline-calibration method can be used. The difference 581 83 Linkping, Sweden, ISBN 91-7373-460-8,
is that the probe of the CMM will measure the Capto 2002
interface of the end-effector. Since this Capto
interface is where the robot holds the end-effector, (5). Kihlman H. and Engstrm H., Affordable
the transform to solve in the Offline-calibration Reconfigurable Tooling, SAE 2002 Transactions
Chuck Journal of Aerospace, ISBN: 0-7680-1285-6, 2002
method is Bushing T . This transform is then feed in the
robot controller as the new tool. The Off-line method (6). Kihlman H. and Loser R., 6DOF Metrology-
is more time consuming than the In-line method, but integrated Robot Control, Aerospace Automated
gives higher accuracy and a more stable environment. Fastening Conference & Exhibition (Aerofast);
September 8-12; Palais des Congrs; Montreal,
Quebec; Canada, 2003
Selecting one of the presented methods must be
decided depending on the production scenario. More (7). Kihlman H., Eriksson I. and Ennis M., Robotic
time available and higher requirements on accuracy, Orbital Drilling of Structures for Aerospace
then the Off-line method is appropriate. In some Applications, SAE Aerospace Automated Fastening
cases however, moving the offset of a pressure-foot, Conference & Exposition, October. 1-3, 2002
would be a waste of time to measure in a CMM, (8). Kyle S., Loser R., Warren D., Automated part
when the metrology system is already available, if the positioning with the laser tracker, Fifth International
accuracy from the metrology system is enough that is. Workshop on Accelerator Alignment, Oct 13-17,
It is a good role of thumb to calibrate in much higher 1997
accuracy than what is needed. Tolerance build-up is
always raising fast. (9). Loser, R., The Authority of Common Measuring
Task Through the Integration of ADM with Laser
Tracker, Coordinate Measurement Systems
6 CONCLUSION Committee, Williamsburg Hospitality House,
This paper presented continuing research on Williamsburg, Virginia, July 8 12, 1996
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
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
2004-01-2830
Magnus Engstrm
SAAB Aerostructures
John Anderson
Advanced Technology Centre, BAE SYSTEMS
This section presents the Dynamic Modules. Some At the top of the Dynamic Modules plate is attached an
parts of the sub-systems presented in this section have interface called a SANDVIK Capto. The Capto interface
previously been presented in [1]. The Dynamic Modules has several purposes, see figure 4. A primary function is
were designed so that an operator could attach them to for the robot to dock onto the modules. A secondary
the static framework manually. Seven different Dynamic function is to attach the pick-ups that are holding the
Modules were developed, see figure 3. 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.
DUMMIES
FINAL REMARKS
Richard Lindqvist
SAAB Aerostructures
Henrik Kihlman
Department of Mechanical Engineering, Linkping University
In a previous paper, a conclusion was made, orbital New future application with a stack of materials
drilling is a suitable method for drilling high quality holes comprising of: 10mm Aluminium + 20mm
in the specific space application that was studied. The Carbon Fibre Reinforced Composite (CFRP) +
conclusion for further development was then to look 10mm Aluminium.
more specifically into the process development, i.e. SAAB Ericsson Space application with a stack
cutting parameters, new cutters, new coatings on of materials comprising of: 0,8mm CFRP +
cutters etc. 18mm Aluminium Honeycombcore + 0,8mm
CFRP
Then the investigation and feasibility study started. The
project went from early prestudy to fully implemented The first part of this paper will present results from the
industrialized solution of the PODU at SAAB Ericsson tests performed in these applications using a PODU
Space (SES) in Linkping, Sweden. More detailed TwinSpinTM 201 that was designed and developed by
description on how the project did it is explained in the Novator AB for SAAB Ericsson Space in Linkping,
paper. Sweden.
The early conclusions made after pre- and final The PODU used at SAAB Ericsson Space was a later
acceptance tests of the PODU at SES shows that version from the predecessor, see figure 1 that was
Orbital drilling is a comprehensive method compared to studied in parallel in the beginning of the ADFAST at
conventionally drilling methods. Today the Orbital drilling SAAB Aerostructures. The unit in figure 1 is the first
method is a certified production process at SES. known developed and manufactured version of the
PODU.
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
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,
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)
RESULTS
Figure 10. The PODU mounted in drill jig ready to
drill holes in Bread Board (BB). The SES application has been studied and the PODU
TM
TwinSpin 201 was later implemented during the year
EXECUTION PHASE 2003 at SAAB Ericsson Space in Linkping, Sweden. In
total, more than approximately 600 holes have been
After the investigation phase the project went in to the drilled as a part of the acceptance and preproduction
execution phase. In this phase the qualification model test program that has been carried out at SES. Only one
for serial production certification was drilled. hole out of the 600 holes that was being drilled in the
QM was detecting a small and single delamination in
The actual acceptance and pre-production test for one layer. The main conclusions and results from this
Orbital Drilling, PODU implementation at SES study could be summarized as follows:
encompassed the drilling of 600 holes in the different
hole sizes in the SES Qualification Model (QM), all holes
Orbital drilling can produce holes within
were drilled within tolerance. Below is pictures taken
tolerance without entry, exit or inter-laminar
from the acceptance and pre-production tests at SES
burrs with PODU machining in these
showed, see figure 11 and 12.
applications, see figure 13.
Pressure-foot
Figure 16: The CNC Robotic Drilling end-effector Our experience and conclusion from this paper
with pressure foot indicates that articulated robots has the following
advantages for drilling:
RESULTS
Articulated Robots are standardized machines
The test specimen was a 13 mm thick aluminum piece,
see figure 17. Hole diameter was inch. Several Being mass-produced for car industry makes
hundred holes where drilled. The end-effector went them cheap in investment
through twice, first a rough cut followed by a fine cut. A
rough cut was performed with a feed-rate of 600 Drill-jigs are replaced by robot programming
mm/min, orbital speed 400 rpm and a spindle speed of
Robots can be re-programmed, which makes ACKNOWLEDGMENTS
production flexible
This work has been a part of the partly funded EU project
Drilling process planning can be done offline, ADFAST (Automation for Drilling, Fastening, Assembly,
saving lead-time in production Systems Integration, and Tooling) and we want to thank all
the people in ADFAST who have supported material to this
An Industrial Robot can be used for other tasks paper. Special thanks go to project manager Mr. John
such as automated fastening, reconfigurable Anderson. Another special thanks go to Novator AB
tooling [7]. 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.
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 REFERENCES
by drilling with Orbital drilling, which produces low
cutting forces. Low cutting force is one key enabler to 1. Latger, F., Harris, T. and Bjrklund S., Drilling Cost
ensure drilling of high accuracy holes. Model, SAE Aerospace Automated Fastening
Conference & Exposition, October. 1-3, 2002.
CONCLUSION 2. Harris, T., Johansson, S. and Ossbahr, G., A Study
of the Influence of Drilling Method and Hole quality
In this paper Orbital drilling implementation and on Static Strength and Fatigue Life of Carbon Fiber
evaluation of PODU TwinSpinTM 201 at SAAB Ericsson Reinforced Plastic Aircraft Material, SAE
Space has been described. Also a description of the Aerospace Automated Fastening Conference &
implementation and evaluation of the CNC-ODU on a Exposition, October. 1-3, 2002.
standard KUKA IRB has been presented. 3. Lindqvist, R., Eriksson I. and Wolf, M. Orbital
Drilling of Sandwich constructions for Space
A global conclusion of what has been presented in this applications, Proceedings of the 2001 Aerospace
paper about Orbital drilling is: Congress (ACE), September10-14, 2001.
4. Helin, P., Jerregrd, H., Robertson, A., and Snell, J.,
"Technologies that make a robot reach Absolut
Orbital drilling is a method capable of producing
Accuracy ", Proceedings of the 33rd International
holes with good hole quality in all the
Symposium on Robotics, October 7-11, 2002.
applications tested
5. Kihlman, H., Eriksson, I. and Ennis, M., "Robotic
Orbital Drilling of Structures for Aerospace
Orbital drilling can give the holes higher surface
Applications ", SAE Aerospace Automated
quality and cleanliness compared to
Fastening Conference & Exposition, October. 1-3,
Conventional Drilling
2002.
6. DeVlieg, R., Sitton, K., Feikert, E. and Inman, J.,
The waste of parts and assemblies due to
ONCE (ONe.sided Cell End effector) Robotic
lacking good hole quality is minimized when
Drilling System, SAE Aerospace Automated
using Orbital drilling
Fastening Conference & Exposition, October 1-3,
2002
Orbital drilling is more difficult to control using
7. Kihlman, H., Ossbahr, G., Engstrm, M., Anderson,
today known handheld equipment, i.e. PODU
TM J. Low-cost Automation for Aircraft Assembly, SAE
TwinSpin 201, because the adjustment and
Aerospace Automated Fastening Conference &
calibration of the machine is sometimes
Exposition, Sept. 20-23, 2004
circumstantial and expensive because the use
of expensive drill jigs, which should be avoided
in the future.
CONTACT
Figure B.1
Figure B.2
Figure B.3
Figure B.4
APPENDIX C
Figure C.1