Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Flexible and Precision Assembly 50.1 Flexible Assembly Automation 885

planner to plan a grasp. Further, they point out that not

all their grasp plans were stable and discuss the need for

further work.

This work further emphasizes the need for a 3-D

modeler integrated with the motion control system to

support both the grasp planning and 3-D object recog-

nition with machine vision.

50.1.3 Flexible Fixturing

Most assemblies are based around a larger part to which

smaller parts are afﬁxed. The larger parttypically moves

through several assembly stations by means of a ma-

terial handling system. Each station feeds and attaches

one or two smaller parts. Some sort of ﬁxture is usually

designed to hold the larger part in a desired orientation.

Current industry practice is for this ﬁxture to be

some sort of pallet, or a two- or three-jaw clamping

system. In some cases ﬁxtured parts must withstand sig-

niﬁcant assembly forces from press ﬁts or secondary

machining operations.

There is interest in reducing the time and cost to de-

sign and fabricate ﬁxtures, and being able to use them

again for a different product rather than scrap them

when a product changes. This has led to various ap-

proaches to ﬂexible part ﬁxturing.

The simplest approach is based on a modular series

of adjustable blocks and clamps that can be used to lo-

cate and support parts. This approach has been used in

machining centers for some time and can provide a high

degree of rigidity.

A more general approach was proposed and tested

by Sandia Labs in 1996. This was a planar clamp-

Fig. 50.7 A computer-generated ﬁxture (after [50.9])

ing scheme in which a geometric planner analyzed the

perimeter of a part, and placed some pins in a loca-

tion on a grid such that a single clamp could uniquely

restrain the part from translation and rotation [50.10].

More recently, a team at Sandia extended this work

to a 3-D planner [50.9](Fig.50.7). Inputs to this plan-

ner include a 3-D (ACIS) model of the part, a ﬁxture

kit speciﬁcation detailing the ﬁxture building tools and

clamps, friction data for the ﬁxture components and the

workpiece, and disturbance forces to be applied to the

ﬁxtured part. The tool then outputs a series of ﬁxture

designs with a quality score for each.

For larger workpieces, one or more multiaxis robots

are now being used to hold and reposition workpieces

while other robots add parts to the assembly. Paral-

lel link structures are being offered commercially for

orienting larger workpieces (Fig.50.8). In November

2007, at the International Japan Robot Exhibition in

Tokyo, Yaskawa showed a two-armed robot perform-

ing assembly, with one arm holding a workpiece and

a second arm inserting components. All programming

for this demonstration was done manually.

Fixture design is essentially the same problem as

grasping design except there is a strong commercial

need for ﬁxtures to be low cost, as for many systems

there may be a large number of ﬁxtures in the sys-

tem. Generating reliable ﬁxture designs quickly from

CAD geometry and knowledge of assembly forces re-

mains an interesting challenge. With the advent of rapid

Fig. 50.8 A robot ﬁxture

Part F 50.1

886 Part F Industrial Automation

prototyping machines, it is now possible to quickly fab-

ricate fairly complex ﬁxture geometry. It is important

that ﬁxtures providekinematic graspsthat canwithstand

assembly (or machining) forces.

50.2 Small Parts

50.2.1 Aligning Small Parts

For many industries part dimensions are shrinking to

a point where humans can no longer handle or assem-

ble parts. Parts with submillimeter dimensions often

have manufacturing tolerances that can be a substantial

percentage of the overall part size, and may not be reg-

istered to any part dimension that can be grasped; for

example, laser diodes emit light with a Gaussian inten-

sity proﬁle whose peak must be measured optically to

position the laser diode in an optical assembly. Cancer

cells in a ﬂuid must be located optically so they can be

sampled by means of a microliter pipette.

While semiautomated systems employing people

looking through microscopes are currently used for

many such applications, advances in integrating ma-

chine vision with motion control now allow parts to

be actively steered into position, using vision to take

a series of pictures to measure alignment error. The

workpiece is moved into position until the alignment er-

ror falls belowa threshold.Commercial applicationsuse

2-D machine vision for this [50.11](Fig.50.9).

In the next several years it is likely that this work

will be extended to 3-D vision. There are some signif-

icant challenges in using 3-D vision for aligning very

small parts. For these applications the parts typically ﬁll

a large portion of the ﬁeld of view of the camera(s).

For 3-D applications this means that, as a gripper ap-

Fig. 50.9 Visual servoing: steering parts into alignment

using vision in motion control loop (courtesy of Precise

Automation)

proaches a part, the part image will change dramatically

in size and perspective. For applications where a num-

ber of pictures will be taken and processed, the vision

system must be able to recognizethese changingimages

without explicit programming for each image. In addi-

tion, for the case where small parts are being actively

mated to each other, one part may obscure the second

part so that only a few features on the second part may

be visible. Lighting, reﬂections, and depth of ﬁeld all

contribute to make this work challenging.

Automatically generating 3-D vision recognition

and part-mating algorithms is another area where the

integration of a 3-D modeling system and a 3-D motion

planning system with the actual online motion control

system will be necessary.

50.2.2 Fastening Small Parts

There are many well-developed techniques for fasten-

ing large parts together. However, when part geometries

shrink to submillimeter levels, fastening techniques are

largely conﬁned to adhesive bonding, eutectic bonding

such as soldering, and welding.

Since all of these material-based fastening tech-

niques involve physical changes in the material, they all

tend to introduce dimensional changes during the fas-

tening process; for example, in aligning laser diodes in

ﬁber-optic transceivers, the desired optical alignment is

within 10 μm. Epoxy curing and laser welding both in-

troduce dimensional shiftsequal tothe desiredassembly

tolerance. It took the photonics industry over 5 years to

improve ﬁrst-pass production yields from 30% to over

90% due to this issue.

There are several approaches to dealing with this

problem. One is to model and predict the dimensional

change that will occur during the bonding process and

offset the assembly positions to account for this. This

only works if the amount of bonding material is highly

repeatable and the assembly geometry is highly repeat-

able. Where the assembly geometry must vary due to

part tolerances, for example where the peak power of

a laser diode moves relative to the package outline,

a ﬁxed offset is not possible. In this case a real-time off-

set would need to be computed based on the alignment

geometry and bonding material properties.

Part F 50.2

Flexible and Precision Assembly 50.3 Automation Software Architecture 887

A second approach is to bond the parts, then mea-

sure the resulting geometry, and deform the resulting

bonded structure to achieve the desired alignment. This

technique is employed by the photonics industry to

align laserdiodes, andis referred to as bendalign. Tech-

niques must be used to stress-relieve the assembly after

deforming so that residual stress does not affect align-

ment over time.

A third approach is to create a separate, rigid, kine-

matic support structure for the parts,which controls part

alignment, and use a parallel bonding process around

this supportstructure toplace itin compression.A prop-

erly designed support structure can then resistthe forces

from the bonding material. The author suggested the

kinematic structure in Fig.50.10, in which a ﬁber is

mounted in either a solid ferrule or a V-block ferrule.

The ferrule rests on a hollow ceramic tube with a V

notch. A solder preform is placed inside this tube. The

assembly system aligns the ﬁber and the preform is

heated, bonding the ferrule to the mounting substrate,

with the ceramic tube resisting the compression forces

when the solder solidiﬁes. A sixth degree of freedom

can beadded by using a half-sphere instead ofa cylinder

for the ﬁber ferrule.

Solder preform

Fig. 50.10 A ﬁve-degree-of-freedom kinematic mount for optical

ﬁbers (courtesy of B.Carlisle)

A fourth approach proposed in the photonics in-

dustry is to bond parts to microservo mechanisms

which can be actively controlled after bonding to align

parts [50.12]. However, this approach is relatively ex-

pensive and may suffer from reliability problems over

extended service periods.

In general, the stable alignment of small parts

to micron or submicron tolerances remains a chal-

lenge, which will beneﬁt from advances in mod-

eling stresses and bonding material deformation in

systems that can be linked to real-time motion sys-

tems.

50.3 Automation Software Architecture

The foregoing discussion points out the need for

some new features in a ﬂexible automation program-

ming language. Some of these features are available

in today’s robot languages; some of them (3-D mod-

eling) are available in separate packages; and some

are not available yet, other than in research environ-

ments. Surprisingly, some robot vendors still do not

offer general-purpose programming languages, prefer-

ring to offer application-speciﬁc programming tools.

For a current state-of-the-art robotic programming lan-

guage, see [50.13]. Also see Chap. 22 on Modeling and

Software for Automation.

A general-purpose automation programming lan-

guage should include the features described below.

50.3.1 Basic Control

and Procedural Features

In addition to common control features such as looping,

branching, multithreading, mathematical functions, and

data structures, an automation language should be able

to coordinate multiple mechanisms on a common time

base. It should be possible to do this over a network,

with a structure that allows master–slave or peer-to-peer

communication and control. It should be possible to use

sensor data to alter motion in real time.

50.3.2 Coordinate System Manipulation

Robots and otherautomation equipment areusually pro-

grammed in a Cartesian coordinate system, which may

be different from the joint coordinates of the robot.

Mathematical models, loosely referred to as kinematics,

transform joint coordinates into Cartesian coordinates

via matrixalgebra. Modern robot languages store Carte-

sian positions in a format known as a homogeneous

coordinate transformation. These coordinate transfor-

mations can be multiplied together to offset positions

relative to a pallet, for example, or a tool. Sensor data

must be transformed from the sensor coordinate system,

for example, a camera frame, to the robot coordinate

system. It is useful to be able to compute elements of

these coordinate transformations in real time at the ap-

plication level; for example, circular, elliptical, spline

Part F 50.3

888 Part F Industrial Automation

or other motions can be computed by a procedure in

a loop that calculates the destination coordinates in a lo-

cation variable and then calls a trajectory generator to

compute and execute an incremental motion. Tracking

a conveyor can be accomplished by using an encoder to

update a base reference frame in real time.

There is increasing interest inincluding a 3-D model

and representationof both robot andworkcell geometry.

Today’s robot languages do not offer the ability to de-

ﬁne the surfaces of the robot or workcell in 3-D space,

other than some simple planes and cylinders. Accurate

representation of 3-D surfaces would be very useful for

collision avoidance, path planning, and safety systems.

There is also increasing interest in improving the

absolute accuracy of robots and other automatic ma-

chines. Today’s kinematic models of six-axis robots

assume that the physical robot is manufactured per-

fectly, i. e., that right angles are perfect, link lengths

are perfect, there are no deﬂections from gravity or

loads, etc. For ofﬂine programming of robots working

in complex environments, for example, spot-welding of

an automobile, it is desirable for the robot motion to be

one to two orders of magnitude more accurate than typ-

ical manufacturing tolerances permit. Recently, more

complete models of robots have been developed in of-

ﬂine simulators. Actual robots can be programmed to go

through a calibration routine and a more accurate model

of the robot is built in the simulator. Then the simulator

commands an offset path which is downloaded so the

real robot will make accurate motions. This is currently

an ofﬂine function. Improved robot controlswill include

this capability as an online function.

50.3.3 Sensor Interfaces

and Sensor Processing

Machine vision, force sensors, and other sensors have

been integrated with commercial robot control systems

since the early 1980s. However, what is needed today is

the ability for sensors to deal with 3-D geometry, and

for sensor programs to be generated automatically from

3-D geometry.

In fact, it would be very useful if sensors could

be used to create 3-D models when no explicit mod-

els exist. A mobile robot entering a new environment

should be able to build up a model of the environ-

ment from sensor data. Extensive work in the research

community has shown success building 3-D models

from a series of 2-D images [50.14]aswellasfusing

sensor data from stereo vision and laser range ﬁnd-

ers for complex navigation tasks such as the Defense

Advanced Research Projects Agency (DARPA) Grand

Challenge competitions in 2006 and 2007, in which

multiple teams of researchers developed vehicles that

could build models and navigate through unknown out-

door environments.

Sensor-generated models must then be available

to the motion planning system so the robot can plan

moves, update progress through the environment, avoid

collisions, and detect and recover from errors.

50.3.4 Communications Support

and Messaging

Many commercially available robot controllers cur-

rently support only limited-bandwidth messaging and

ﬁle handling. With the rapid development of distributed

motion control where motion can be coordinated at

frequencies of 1 kHz or more over a network, the

technology now exists for higher-bandwidth, time-

synchronized communication; for example, a mobile

camera could broadcast an image to a group of mo-

bile devices (imagine soccer-playing robots) that also

exchange strategy and planning information. The au-

tomation language and operating system should be

capable of handling high-bandwidth communications

without interfering with other deterministic tasks such

as trajectory planning, image processing, and servo

loops. The ability to set time slots and task priorities in

the operating system becomes very important as com-

munication loads increase. Personal computers remain

disappointing in this regard, with large, unpredictable

communication delays.

50.3.5 Geometric Modeling

From much of the foregoing discussion it should be

clear that the author believes that automation languages

should be extended with 3-D modeling systems. One

important difference though, is that robotic and other

automation systems are dynamic, while many model-

ing systems are static, or updated at a low rate. In order

to be useful, an online modeling system needs to be

capable of being updated at rates similar to those in

3-D video games, as robots can make large motions in

a few hundred milliseconds. It is likely that some of

the simpliﬁcations and data-compression methods used

in video games may be useful for real-time geometric

modeling for motion control.

Eventually, it may be useful to include dynamic

models as well as structural deﬂection models in au-

tomation languages. Today’s robots are large, rigid,

Part F 50.3

Flexible and Precision Assembly 50.3 Automation Software Architecture 889

heavy structures with masses that exceed their rated

payloads by a factor of ten or more. Lighter, more ﬂexi-

ble robots would waste less energy, but will be harder to

control and harder to program without some capability

to predict their trajectories under load.

50.3.6 Application Error Monitoring

and Branching

In most industrial robot applications, over 50% of the

programming effort is devoted to anticipating, sensing,

and recovering from possible errors. Today, much of

this programming is done on site, in an ad hoc man-

ner, and takes a long time to develop and debug. As

more sensors are used, this problem becomes larger, as

sensors can introduce new errors. It may not be obvi-

ous to the programmer why a sensor-driven system is

not reliable. To aid debugging, data-logging features,

time-stamping of data and communication messages,

and single stepping of motion programs are becom-

ing common automation language features. In assembly

systems, many errors are due to poorly understood or

poorly modeled part tolerances. Often weeks or even

months of testing is required for a production system

to meet reliability standards. This testing is used to ﬁnd

software bugs, make sensors reliable, and make the sys-

tem robust within a statistical range of part tolerances.

However, we still do not have general methods to

analyze assembly systems for errors, represent errors or

generate error monitoringand recovery strategies. Work

by Deming and others in statistical process control con-

tributedgreatly to understandingmanufacturing process

tolerances and designing products and processes for

reliable production with known tolerances. However,

these techniques, while used in the metal-forming and

semiconductor industries, seem to be largely absent

in automated assembly systems. In 1997 Carlisle and

Craig developed a simulation tool [50.15] for assem-

bly tolerance process analysis that was used by Nokia

to analyze and improve cellphone production yields.

However, since assembly systems vary so dramatically,

there are few, if any, generally accepted practices for

modeling assembly processes and part tolerances and

predicting and improving yields.

It is not clear if it is productive to try to predict er-

rors. In general, an error is a deviation from a plan. It

may be more productive to detect errors quickly and for

the system to have enough geometric and sensor data to

make a new plan quickly to try to recover from the er-

ror. Assembly systems with this capability have yet to

be demonstrated.

50.3.7 Safety Features

The cost of assembly robots and sensors is coming

down quickly. At the same time the speed of these

devices is increasing. Motions of 1 m in a fraction of

a second are now common. As a result, robots can

present a substantial danger to humans who may enter

the robot workspace.

The industry approach to this issue is to create walls

around the robot with sensors and interlocks to prevent

people from entering the robot workspace when it is

moving under computer control. This approach is both

expensive and inefﬁcient. A US$15000 robot may be

surrounded by US$5000 of screens, light curtains or

safety mats. Creating cells with walls tends to require

more ﬂoor space for each cell.

More generally, there are more and more applica-

tions where it is desirable to have robots work with,

and in some cases touch, people. To address this is-

sue in a general way we need control systems that can

model the robot’s structure as well as the environment,

and sensors that can detect people entering the robot’s

workspace. We need motion control systems that can

respond dynamically to space intrusion and modify the

motion appropriately. An operator should be able to

walk up to a robot workcell and load a new tray of

parts in the workspace without fear of injury in the

same manner that he or she wouldinteract with a human

assembler.

50.3.8 Simulation and Planning

It is time for robot simulation to move from an of-

ﬂine capability to an online capability, where the

control system contains a real-time geometric simu-

lation of the complete assembly system. Simulation

systems are now widely used for programming robots

for spot-welding, arc-welding, and some material han-

dling tasks. However, they could be more widely used

for many applications described here, including pro-

gramming ﬂexible part feeders, programming 2-D and

3-D vision systems, optimizing sensor-driven motions

through workcells, detecting and recovering from er-

rors, and allowing robots to interact safely with people.

Online simulation offers the opportunity to develop

high-level representations of common tasks; for ex-

ample, a task-level command such as “Drive a screw

at location hole 1 to torque X” is much easier for

an application programmer to work with than many

lines of detailed programming code. However for task-

level instructions to be fairly general, they should be

Part F 50.3

890 Part F Industrial Automation

able to access geometric and process information from

databases, and a simulation system with motion plan-

ning ability and knowledge of the robot and workcell

should generate the motion plan for the task. If an error

occurs, the planning system should generate a new plan

online.

More generally, simulation and task-level planning

are necessary tools to address the broader issue of robot

interoperability, and program and data sharing. Users

would like to be able to move programs from one brand

of robot to another. This will remain an elusive goal

until robots can be instructed at a very high level of ab-

straction, in almostthe same manneras a humanis given

high-level, abstract directionsalong withsome data,and

ﬁgures out how to perform the task. The instruction

“Bolt down the manifold cover”, given a CAD model,

is far easier to translate than today’s explicit low-level

programs.

50.3.9 Pooling Resources and Knowledge

In the last few years it has become technically pos-

sible for people to share knowledge and collaborate

remotely over the Internet.This hascreated thepotential

for open-architecture systems where many people pool

knowledge to create complex systems efﬁciently. The

Linux operating system is one example that has been

widely accepted by an enthusiastic user base.

An ideal robot programming system would have

a core set of functions that could support a wide range

of applications, yet it would be open enough and ex-

tensible enough that users could create new capabilities

that could be shared or integrated. There is an inter-

esting balance between providing enough structure that

programs from different developers can be integrated

while allowing enough ﬂexibilitythat users can add new

features easily.

50.4 Conclusions and Future Challenges

In general,we need to raise the abstractionlevel ofrobot

programming if robots are going to be able to perform

increasingly complex tasks in increasingly less struc-

tured environments. Programming languages for robots

need to incorporate modeling, sensing, and planning ca-

pabilities to allow libraries of tasks and actions to be

compiled. These high-level tasks need to be robust and

safe. We also need to think about how to pool ideas and

resources frommany different developersand locations,

to build up the large knowledge base that will be needed

for robots to move from simple, highly structured tasks

to complex, unstructured tasks. A more thorough dis-

cussion on the future of ﬂexible and precise automation

is provided in Chap. Sect.21.3 of the handbook.

50.5 Further Reading

1. S. Y. Nof (Ed.): Handbook of Industrial Robotics

(Wiley, New York 1999)

2. B. Siciliano, O. Khatib (Eds.): Springer Handbook

of Robotics (Springer, Berlin, Heidelberg 2008)

References

50.1 A. Shaﬁ: Bin Picking Axle Shafts (Shaﬁ Inc., Michi-

gan 2008), http://www.shaﬁinc.com/solutions/sol36/

test_1.htm (last accessed 2008)

50.2 K. Goldberg, B. Mirtich, Y. Zhuang, J. Craig,

B. Carlisle, J. Canny: Part pose statistics: estima-

tors and experiments, IEEE Trans. Robot. Autom.

15(5), 849–857 (1999)

50.3 B. Carlisle: Feeder developed by at Adept

Technology and licensed to Flexomation Inc.

http://www.ﬂexomation.com/ (last accessed 2008)

50.4 Flexfactory: Feeder developed by Flexfactory AG

(Flexfactory, Dieticon 2008),

http://www.ﬂexfactory.com/ (last accessed 2008)

50.5 A. Miller, P. Allen: From robotic hands to human

hands: a visualization and simulation engine for

grasping research, Ind. Robot 32(1), 55–63 (2005)

50.6 Barret Technology: http://www.barrett.com/robot/

products-hand.htm (2008) (last accessed 2008)

50.7 Schunk Gripper: http://www.schunk.co/ (2008)

(last accessed 2008)

Part F 50

Flexible and Precision Assembly References 891

50.8 D. Krajic, A. Miller, P. Allen: Real time tracking

meets online grasp planning, Proc. 2001 ICRA IEEE

Int. Conf. on Robotics and Automation, Vol. 3 (2001)

pp. 2460–2465

50.9 R. Brown, R. Brost: A 3-d Modular Gripper Design

Tool, Sandia Rep. SAND97-0063, UC-705 (1997)

50.10 R. Brost, K. Goldberg: A complete algorithm for de-

signing planar ﬁxtures using modular components,

IEEE Trans. Robot. Autom. 12(1), 31–46 (1996)

50.11 J. Shimano: Visual Servoing Taylor Made for

Robotics (Motion Systems Design, 2008), (last ac-

cessed 2008)

50.12 A. Hirschberg: Active alignment photonics assem-

bly, US Patent 6295266 (2001)

50.13 B. Shimano: Guidance Programming Language

(Precise Automation, 2004–2008),

www.preciseautomation.com (last accessed 2008)

50.14 M. Lin, C. Tomasi: Surface Occlusion from Lay-

ered Stereo. Ph.D. Thesis (Stanford Univ., Stanford

2003)

50.15 J. Craig: Simulation-based robot cell design in

AdeptRapid, Proc. 1997 ICRA IEEE Int. Conf. on

Robotics and Automation, Vol. 4 (1997) pp. 3214–

3219

Part F 50

“This page left intentionally blank.”

893

Aircraft Manu

51. Aircraft Manufacturing and Assembly

Branko Sarh, James Buttrick, Clayton Munk, Richard Bossi

Increasingly the manufacturing of complex prod-

ucts and component parts involves signiﬁcant

automation functions. This chapter describes

a cross section of automated manufacturing sys-

tems used to fabricate, inspect, and assemble

aircraft. Aircraft manufacturing cost reductions

were made possible by development of advanced

technologies and applied automation to produce

high-quality products, make air transportation af-

fordable, and improve the standard of living for

people around the globe. Fabrication and assem-

bly of a commercial aircraft involve a variety of

detail part fabrication and assembly operations.

Fuselage assembly involves riveting/fastening op-

erations at ﬁve major assembly levels. The wing

has three major levels of assembly. The propulsion

systems, landing gear, interiors, and several other

electrical, hydraulic, and pneumatic systems are

installed to complete the aircraft structurally and,

after functional tests, it normally gets painted

and goes to the ﬂight ramp for ﬁnal customer

acceptance checks and delivery. Aircraft manufac-

turing techniques are well developed, fabrication

and assembly processes follow a deﬁned se-

quence, and process parameters for manual and

mechanized/automated manufacturing are pre-

cisely controlled. Process steps are inspected and

documented to meet the established Federal

Aviation Administration quality requirements,

ensuring reliable functions of components, struc-

tures, and systems, which result in dependable

aircraft performance.

51.1 Aircraft Manufacturing

and Assembly Background..................... 894

51.2 Automated Part Fabrication Systems:

Examples ............................................. 895

51.2.1 N/C Machining

of Metallic Components................. 895

51.2.2 Stretch Forming Machine

for Aluminum Skins ...................... 897

51.2.3 Chemical Milling and Trimming

Systems for Aluminum Skins .......... 898

51.2.4 Superplastic Forming (SPF)

and Superplastic Forming/

Diffusion Bonding (SPF/DB) ............ 899

51.2.5 Automated Composite Cutting

Systems ...................................... 900

51.2.6 Automated Tape Layup Machine..... 901

51.2.7 Automated Fiber Placement

Machine...................................... 902

51.3 Automated Part Inspection Systems:

Examples ............................................. 903

51.3.1 X-ray Inspection Systems .............. 903

51.3.2 Ultrasonic Inspection Systems ........ 904

51.4 Automated Assembly Systems/Examples.. 905

51.4.1 C-Frame Fastening Machine .......... 906

51.4.2 Ring Riveter

for Fuselage Half-Shell Assembly.... 906

51.4.3 Airplane Moving Line Assembly ...... 907

51.5 Concluding Remarks

and Emerging Trends............................ 908

References .................................................. 909

The emergence of the industrial age brought sig-

niﬁcant changes and impacts on living conditions

of human societies. Innovative product development

in a free market environment, driven by a desire

to improve standards of living, led to revolutionary

products in computing, communications, and trans-

portation that impacted all levels of human activities.

The key enabler of this progress was improved in-

dustrial productivity, which climaxed with automation

technologies, starting with hard automation (suit-

Part F 51

894 Part F Industrial Automation

able for mass production of consumer goods) and

evolved into intelligent automation (selectively used for

batch-type single-component fabrication) using com-

puters and software to precisely control complex

processes.

Automation has successfully met the need to im-

prove quality, reduce cost, and improve ergonomics

of aircraft fabrication and assembly. The conversion

to digitally deﬁned aircraft and advancements in ma-

chine tools have enabled widespread use of automation.

The trend to machine/fabricate components accurately

with automated machines started with a Massachusetts

Institute of Technology (MIT) demonstration of the

ﬁrst-ever developed numerically controlled (N/C) ma-

chine in 1952. This quickly led to a huge machine tool

market that enabled rapid production of precision ma-

chined parts and accurate assembly of large aircraft

structures.

51.1 Aircraft Manufacturing and Assembly Background

Aircraft manufacturing cost reductions were made pos-

sible by development of advanced technologies and

applied automation to produce high-quality products,

make air transportation affordable, and improve the

standard of living for people around the globe [51.1–3].

Fabrication and assembly of a commercial aircraft,

such as the one depicted in Fig. 51.1, involves a vari-

ety of detail part fabrication and assembly operations.

A number of raw materials are machined and fabricated

into detail parts, which are then assembled into various

levels of structural conﬁgurations. Starting with basic

assembly of detail parts into simple panels,they are then

combined into super panels and higher-level assemblies

to produce the fuselage, wings, and ﬁnally the com-

plete aircraft. Integral designs and efﬁcient production

of new aircraft involve tradeoffs to optimize materials,

the number of parts and size of structures, use of inno-

vative processes,and adaptations of appropriateexisting

equipment and facilities.

Fig. 51.1 Major aircraft fuselage and wing components layout

Fuselage assembly involves riveting/fastening oper-

ations at ﬁve major assembly levels. At the ﬁrst (lowest)

level, skins, doublers, longerons, and shear ties are

joined to form a single panel. The size and complexity

of single panels is primarily driven by aerodynamics,

load requirements, and function in operation of the

aircraft. At the second assembly level, several single

panels are joined using additional detail parts along lon-

gitudinal and radial joints into super panels. Frames

are usually attached to shear ties during this operation.

Assembly of the ﬂoor grid structure (also second-level

assembly) joins the ﬂoor beams and seat tracks. At the

third assembly level, half shells are created by joining

super panels and the ﬂoor grid is often attached to the

upper or lower half-shell of the fuselage. Fourth-level

assembly involves longitudinal joining of barrel halves

to complete the individual barrel structures. Fifth-level

(highest) assembly does the 360

◦

radial joins which fas-

ten the nose, forward, center, and aft fuselage barrels

together. Inside these structures, multiple radial dou-

blers, couplings, and ﬁttings are installed to complete

the aircraft’s fuselage structure.

The wing has three major levels of assembly. The

ﬁrst (lowest) assembly level joins the upper and lower

skin panels, spars, and bulkheads (consisting of N/C-

machined skins, stringers, and stiffeners). At the second

assembly level, spars and bulkheads are joined to form

the wing grid,to which skinpanels areattached to create

the wing box. The third (highest) assembly level joins

leading and trailing edge components to the wing box

to complete the wing structure. The wing is joined to

the fuselage in the appropriate sequence to complete the

airframe.

The propulsion systems, landing gear, interiors, and

several other electrical, hydraulic, and pneumatic sys-

tems are installed to complete the aircraft structurally

and, after functional tests, it normally gets painted and

Part F 51.1