HD28
.M414
9/
3&*
mm*
ALFRED
P.
^v
7:
11 iwi
N
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
Cognitive Complexity and CAD Systems:
Beyond the Drafting Board Metaphor
David Robertson
Karl Ulrich
Marc Filerman
WP#
3238-91-MSA
January 1991
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
Cognitive Complexity and CAD Systems:
Beyond the Drafting Board Metaphor
David Robertson
Karl Ulrich
Marc Filerman
WP#
©
3238-91-MSA
January 1991
1991 Massachusetts Institute of Technology
Sloan School of Management
Massachusetts Institute of Technology
50 Memorial Drive
02139
Cambridge,
MA
Acknowledgement
The research described in this paper was performed at the MIT Artificial Intelligence
Laboratory and at the MIT Sloan School of Management. Support for the
laboratory's research is provided in part by the Advanced Research Projects Agency
of the Department of Defense under Office of Naval Research contract N00014-85-K0124. Additional support for this research was provided by the MIT Leaders for
Manufacturing Program, an educational and research partnership between MIT and
eleven major U.S. corporations. In addition to the authors, Daniel Berkery and
Jeremy Yung were members of the team that performed this work.
M
IT. LIBRARIES
MAR
1
1
1991
ABSTRACT
Computer- Aided Design (CAD) systems can and should support and enhance the
product development process.
Unfortunately, the benefits delivered by current
We believe that CAD systems should be
systems have not met users' expectations.
designed to minimize the cognitive complexity facing the engineer; CAD systems
should be easy to use and should help the engineer manage design-related
complexity.
A series of propositions are developed which refine these ideas.
To evaluate
the central propositions,
we constructed
a prototype
CAD
system for the
The user of the system is provided
design of blanked and bent sheet metal parts
with a hand-held input device which interprets actions of the user's hands as
.
production operations on a
CAD
representation of the part.
The user creates sheet
metal parts by bending, stretching, pushing, and moving the input device.
The system was demonstrated
who provided
to
engineers, engineering managers, and researchers,
ideas for future enhancements.
Reactions to the demonstrations of
Although we
have used sheet metal as an example domain, we believe these ideas can be applied
in a broad array of design contexts.
the system have helped evaluate
the concepts
behind the system.
1.
Introduction
Computer-Aided Design (CAD) systems should support and enhance
development process. Given the complexity of product development,
which may
The complexity
conflict.
of the product
cognitive complexitv for the engineer;
amount
in
many
Product development entails balancing
difficult task.
this is a
constraints,
some
of
development process leads
must mentallv juggle
to
a large
This cognitive complexity can lead to a situation
of important information.
which important design
the engineer
the product
factors are not considered
and
The
failures result.
consequences of such failures can be economically punishing or even tragic (such as
the Kansas City Hyatt disaster [Petroski 1985]). Unfortunately,
available
CAD
manage
systems add
to the
model
CAD
a design.
Indeed, because they are hard to use,
that complexity.
complexity facing the engineer.
of the design
modify
and the
CAD
system. This
Yet for
many
Our
command
syntax
command
to create or
and
idiosyncratic
is
and memorization
to learn
and use
Report 1990).
central
argument
is
that
CAD
systems should help minimize the cognitive
complexity facing the engineer. More specifically,
should
1) assist in
accomplish
we
argue that
managing design-related complexity and
this goal,
• utilize
To
the
is
syntax must be manipulated
systems, this
many
Between the engineer's mental
representation of that design
command
unintuitive, requiring a great deal of training
(CAD
of the currently
systems do not reduce the complexity facing the engineer nor help
the engineer
syntax of the
many
we argue
that
CAD
a production-like design
2)
CAD
be easy
to use.
To
systems should:
metaphor,
•
evaluate designs with respect to the performance criteria,
•
embed
•
automate well understood design
•
incorporate natural mappings in their controls and display.
test
systems
well understood geometrical constraints,
our ideas,
we have
tasks,
built a prototype
and
CAD system
bent sheet metal parts. The user of the system
is
for designing
blanked and
provided with a hand-held input
device that interprets the three-dimensional actions of the hands in real time as
production operations on a visual display of the part. Using the system, the user
creates sheet metal parts
device.
by bending,
stretching, pushing,
and moving the input
In the four remaining sections of this paper,
design and
CAD systems
we provide background on
technology, propose a set of ideal
characteristics, describe a
CAD
we
system
ive
built,
CAD
engineering
system
and discuss the
results
and
implications of our research.
2.
Background: Product Development and
CAD Systems
Design can be thought of as a multiple objective, constraint satisfaction problem.
must achieve many diverse performance
Typically, successful designs
satisfy
targets
and
competing constraints. For example, designing a gas turbine engine may
involve meeting customer-specified performance criteria such as thrust, fuel
consumption,
cost,
and weight. Individual engine parts may have design
and constraints such
The part should be
as rigid as possible.
•
Stiffness.
•
Stress.
•
Temperature
•
Geometry. The part must attach
•
Producibility.
at the
objectives
as:
The part should not
resistance.
fail
under the design
The part should have
load.
satisfactory material properties
operating temperature.
to a specified interface
and not
interfere with
adjacent parts.
Note
The
part should be producible, preferably
that in general there
criteria to
be traded
is
no overall objective function
existing equipment.
that allows
how much
For example,
off against another.
on
one of the
production cost
is
another unit of stiffness worth? Also note that some of the constraints are firm and
can not be violated under any circumstances
must never exceed the
(for
example, the stress on the part
yield stress) while others are soft
necessary (parts that cannot be
made
and can be violated
if
with existing production equipment can be
supplied by a vendor).
Given the number and complexity of the
is
designs optimized along
to
criteria
and
constraints, the design process
usually one of satisficing (Simon 1981) rather than optimizing. Only rarely are
more than one
or
two
myopically improve the design with respect
the constraints until the design
[1984] for
an example of
this).
is
satisfactory
Much
criteria.
to
Rather, the usual process
one constraint, then
from
all
(see Calkins
of the justification for
CAD
is
iterate across
and Ishimaru
system purchases
is
based on potential improvements
to the
well-known
Many
design managers hope that
process that will
2.1
CAD Systems
Many
articles
make
it
CAD
just
by CAD.
systems will enable changes in the design
less iterative, less
have detailed
to
design process, not
myopic, and more productive.
Use
from the application
been shown
to the effectiveness of this
efficiency gains in particular tasks that are enabled
of
have arisen
stories of large productivity benefits that
CAD
technology
to
engineering work.
CAD
systems have
reduce design time (Fitzgerald 1987, Bull 1987, Teresko, 1988, 1990,
Manji 1989, Frangini 1990), reduce design costs (Smith 1982, Dutton 1986, Fitzgerald
1987, Lansiaux 1987,
Krouse
et
al.
1989),
and improve design quality (Crombez
Eade 1988, Vasilash 1988, DeMatthew 1989, Velloci and Childs
literature also details stories of
unmet expectations and daunting
successful deployment (Majchrzak
to revolutionize the
and Salzman
1989).
Problems have included software defects, slow
1990).
the technology are quite
mixed
leads to the conclusion that a
a
most current
that
CAD
of
use
is
by
of drafters serving a
CAD
systems 1
But the continued use of
due
CAD
of their time
systems are
(CAD
CAD
literature
.
(CAD Report 1990).
Many companies have traditionally had
group of engineers; the engineer would sketch or
detailed drawing.
most 25%
of the
drafters, not engineers
otherwise specify the design geometry for the drafter,
to the fact that
Manager's opinions of
ergonomic and organizational issues have
true partially for historical reasons.
group
Any review
(Farrar 1987).
number
hindered the successful deployment of
is
obstacles to
Systems that were expected
learning curves, and complex user interfaces (Salzman 1989).
This
1988,
Yet the
product development process delivered only meager
improvements (Adler
Note
1990).
CAD
who would
then produce a
systems primarily by drafters
difficult to learn
and
use.
Engineers use
Report 1990). The memorization of
is
also
CAD
at
CAD commands
many situations a CAD specialist (e.g. drafter) is
(CAD Report 1990). We believe that engineers could
effectively use CAD systems much more than they do, but mastering the systems is
currently too time-consuming a task for many engineers.
and syntax
needed
1
is
so difficult that in
to operate the
system
See, for example, the special issues of
devoted
to
CAD
/£££ Transactions on Engineering Management
systems (August and November 1989).
We
argue that the
difficulties
are partially rooted in
designed
CAD
to
systems are
difficult to
recent advances in
CAD
Advances
in
use for design.
CAD
systems technology
technology
We
are not the only researchers and
in these
is
a
two problem
review of some
areas.
Systems Technology: Supporting Engineering
Several recent advances have allowed
We
CAD
systems have historically been
have noticed these problems. Following
to
work.
CAD
support the drafting function, not engineering design. As such, most
system developers
2.2
experienced in effectively deploying
two major problems:
CAD
discuss three such advances:
systems
to better
support engineering
rule-based parametric design, integration of
design and analysis packages, and feature-based design. Voelcker and his colleagues
(1988) provide a
The
first
comprehensive survey of these and related technologies.
of these, rule-based parametric design, allows the
part geometry for specific families of parts.
Several
CAD
tools that allow engineers or tool developers to integrate
commands with
automated generation
of
vendors have developed
geometric modeling
other types of problem-solving tools like object-oriented
programming languages and rule-based systems. These
development of special-purpose
tools
have enabled the
tools for particular types of design tasks.
may
example, a firm that manufactures cutting tools
develop a
CAD
For
application that
automatically produces a tool design from a specification of the required removal
rate,
geometric constraints, workpiece material, and coolant.
The second advance,
better integration of design
and analysis packages,
lets
the user
quickly analyze the mechanical or thermal stresses on a part, as well as the
kinematic behavior of a part. Improved interfaces between design and analysis
packages make such analysis possible with the original part geometry and relatively
little
extra work.
This allows the engineer to perform analyses on
many more
design variations.
Feature-based design provides the engineer with design primitives based upon part
features such as holes, slots, bosses, or flanges.
composed
These primitives are typically
of several lower-level geometric primitives
and so increase the speed
with which part geometry can be specified. The features are also easier to remember.
Further, the features often correspond to manufacturing processes, so that the
design of a part more closely matches the production of that part.
2.3
Advances
in
CAD Systems Technology: CAD
Interfaces
There has been some research on improving the interface between the user and the
CAD
Pentland (1987) developed a sketching tool called "Super Sketch"
system.
which allows users
The
virtual
slider bars.
"lumps"
to
lumps
to create visual
images by deforming "parametric lumps of clay"
and deformed using
of clay are manipulated
a
mouse
change
to
The user can perform Boolean operations on an arbitrary number
Other research
efforts
have focused on input devices
shared by these researchers
is
that creating
Polhemus 3Space input device2
.
for
CAD
systems. The premise
and manipulating three-dimensional
images requires three dimensional controls.
Several teams have used the
This device uses low frequency magnetic waves to
provide the three-dimensional position and orientation of a sensor relative
stationary source.
VPL
mounted display device
world.
to
.
The "Eyephone"
that provides the user with a stereoscopic
The Polhemus sensor
tracks
in the virtual
a user's
hand
system
to
this
images
hands
(Schmandt
in a virtual stereoscopic
Polhemus,
3
VPL
Inc.,
Research,
their
a glove
When
used together,
with a virtual world.
3-Draw (Roberts
et al. 1986)
Redwood
It
has
objects.
1983).
world that
Colchester VT.
Inc.,
is
is
in a
"Magic Wand" that
With
his
is
held in
system a user can paint 3-D
projected into a space in front of the
Other systems which allow sketching and visualization
three dimensions are the
Workstation (Waldern
2
to interact
Polhemus sensor
a
like a pencil
user by two monitors.
world with
kind of interface could be used in conjunction with a
manipulate three-dimensional
Schmandt has incorporated
the user's
of each of the joints in the hand.
and DataGlove" enable users
been hypothesized that
CAD
a virtual
that produces signals corresponding to the position of the
and the angles
the "Eyephone"
a head-
view of
The "DataGlove"
absolute position and orientation in the real world.
user's wrist
is
3-D head movements and enables the computer
match the viewer's position and orientation
worn on
to a
Research in California has incorporated Polhemus devices
"Eyephone" and "DataGlove" products 3
in their
of
obtain the desired geometry.
City,
CA.
1989)
systems.
and the
Interactive Graphics
in
Finally,
NASA
has developed the 3-D "Sensor Frame" input device that senses
and orientation
finger position
(NASA
1989).
volume
in a
in front of the
CAD
This device could be used as a
computer monitor
input device that allows
designers to "sculpt" virtual parts simply by moving their hands.
3.
Propositions for
Our
goal for
CAD
CAD System Design
system design
We
engineering tasks.
CAD
system
to
minimize the cognitive complexity
suggested in the introduction that
from the inherent properties
the
is
we
research, develop five propositions for
CAD
and system complexity-
first
characteristic influences task
premise
Humans have
is
that engineers
need
from the use of
present the premises underlying our
proposed
Our
complexity arises
of engineering design tasks as well as
In this section
itself.
this
of
system design, and discuss the
way
each
a tool to assist in creating part geometry.
severely limited short term
memory
limitation contributes to the sequential
and
satisficing character of design solutions
(Simon
capacity (Miller 1956).
iterative nature of
1981).
design and
This
to the
Designers would be unable to
design even the simplest parts without some form of external working memory.
This
was
true long before the advent of computers: drawings, sketches,
have been used
to facilitate
Our second premise
does not have
is
that
to
(McKim
design problem solving for centuries
computers
will
be used
to
implement
The external working memory required
part geometry.
and models
1980).
tools for defining
for design
problem solving
be stored, displayed, and manipulated by computer. (Many
automobiles were designed with pencil and paper as recently as ten years ago, and
some gas
need
turbine engines are
to transmit
and
still
designed on drafting boards.) Nevertheless, the
distribute design data to
many
people, to store and reproduce
designs, to evaluate and analyze designs using computer programs,
incrementally edit designs
tools
Our
make computers
and
to
the likely vehicle for design support
(Majchrzak and Salzman 1989).
final
premise
is
that
dimensional view of the
dimensional display
(as
CAD
systems will display images representing a three-
part.
Researchers have
opposed
to several
shown
that a
system with a three-
two-dimensional views) increases the
performance of users carrying out tasks which require the perception and
understanding of spatial information (Bemis
et al. 1988).
This
is
especially true
when
the displays
show
frame images (Barfield
3.1
CAD
surfaces, as surfaces are
Systems Should Utilize
a Production-Like
The operating metaphor underlying
a
CAD
system
primitives and operations presented to the user.
metaphor
utilized the
of the drafting board:
manipulating lines and arcs
pencil
on
more
easily interpreted than wire-
et al. 1988).
much
in
the
Design Metaphor
is
defined by the choice of
Historically,
with lines and arcs as a metaphor
systems have
drafters create part designs
same way
as they are
There are several drawbacks
a drafting board.
CAD
for constructing
to
by
manipulated with
a
two-dimensional drafting
three-dimensional part geometry.
One problem is the detail of the primitives; constructing a simple block with a hole
through it may require dozens of lines and arcs. Another problem is that lines and
enough expressiveness
arcs provide the designer with
even physically
to three
arcs.
to create parts that are
Some commercial systems extend
realizable.
dimensions by allowing wire-frame images
to
the drafting
not
metaphor
be constructed with
lines
and
These systems suffer many of the problems of two dimensional systems.
Constructive solid geometry modelers, which are designed around geometric
primitives and Boolean operations on those primitives, provide the clear benefit of
enforcing physical realizability-
drawbacks
for designing parts
However, these modelers have some serious
with complex curved surfaces (Pratt 1984, Computer
Graphics World 1988).
We believe that CAD systems should utilize the metaphor of a production process.
A CAD system designed around this metaphor utilizes commands that correspond
to production-like operations.
For example,
when
designing a machined part, the
engineer would use a milling-like operation to create a slot (Cutkosky and
Tenenbaum
1987).
types of parts, e.g.
A
strict
it is
squirting liquid plastic
shape processes
may
production process metaphor
is
not appropriate for
all
not dear that providing the engineer the capability of
would help design
(e.g. casting,
injection
molded
parts.
In the case of net
molding, or forging), the tooling fabrication processes
be more appropriate choices for design operations.
The production metaphor
and ease of
use.
are likely to be
offers several
advantages
in
both supporting design tasks
Parts designed through a series of manufacturing-like operations
more producible than those designed through
traditional
CAD
systems, freeing the engineer from constantly assessing the production implications
make
CAD
Designing a
of the design.
the systems easier to learn.
system around a production metaphor
The commands are more
easily
will also
remembered,
as
they relate to physical operations such as mill, bend, and cut.
Both constructive solid geometry modelers and systems utilizing production process
metaphors
may
in
limit the geometrical expressiveness of the engineer.
some
cases be inconvenient.
For example,
when
design paradigm would prevent designing a part by
We
sectional view.
argue that
this
inconvenience
is
first
This limitation
interpreted rigidly, this
constructing a cross-
overshadowed by
the
advantages the metaphor provides.
3.2
CAD Systems Should Evaluate Designs
many
Successful part design requires satisfactory performance along
dimensions.
For even the simplest bracket, the criteria include cost, weight, stiffness, strength,
and geometric
interference.
Rather than requiring engineers to constantly evaluate
designs with respect to the performance
display as
much
problem solvers
example, when
criteria,
of this information as possible.
if
CAD systems should compute and
We believe that engineers are good
given adequate performance feedback on their designs. For
designing an inner body panel for an automobile, a designer could
be presented with the number of required dies, the cost of the tooling, the cost of the
material for the part, the
number
of welds required to attach the part to
its
neighboring parts, and the contribution of the part to the torsional and bending
This feature of
stiffness of the vehicle.
CAD
systems would serve primarily
the cognitive complexity facing the engineer.
to certain criteria
would be made by
to ease
Evaluations of the design with respect
the system,
and would be
easily available to the
engineer.
3.3
CAD
Systems Should Embed Well-Understood Constraints
Another way of reducing task complexity
procedures of the
CAD
is
to
system. For example,
if
embed design
a pocket
is
constraints in the
being created
in a
block of
material, the internal corner radii could be constrained to be greater than the
minimum
diameter end mill allowable for that operation.
constrained to be far enough
fit.
away from
a wall that the corresponding bolt
In the simplest implementation of this idea, the
allow certain operations
if
Similarly, holes could be
CAD
head
will
system simply does not
they violate specified constraints.
If
the procedures of the
CAD
system thereby limit the degrees-of- freedom available
to the engineer, the
engineer can focus more attention on other issues.
The idea
embedding well understood
of
constraints in the
from but compatible with the idea of designing the
production metaphor.
Many
constraints that could be
production operations, or are related
For example, constraints on
CAD
how
to a single
CAD
system
is
distinct
system around a
embedded
are not related to
company's production
close holes can be placed to the
edge
capability.
of a sheet
metal part are different for laser-cut sheet metal than for blanked sheet metal. Thus
a
CAD
system designed around a production metaphor should not enforce
The implementation of the
constraint.
may embed
3.4
CAD
CAD
system
at a certain
this
company, however,
such constraints.
Systems Should Automate Well Understood Design Tasks
Many commercial CAD packages provide programming
CAD commands to be chained together. In
sequences of
languages that allow
this
way
well understood
design tasks can be automated. Rule-based parametric design systems allow the
automatic generation of part geometry from a series of design rules and basic
These design rules can include rules relating
parameters.
to material properties,
weight and cost minimization, production process limitations, and other
constraints.
Automation of design
concentrate on
3.5
CAD
more
critical
tasks,
however
it is
done, frees the engineer
to
design issues.
Systems Should Incorporate Natural Mappings in their Controls and
Display
There appear
be strong similarities between the cognitive and the physical
to
manipulations of complex three-dimensional objects (Shepard and Metzler 1971,
Shepard and Feng
1972).
For example,
if
asked
to
compare two pictures
of three-
dimensional objects, a person will mentally "rotate" one of them in a manner
analogous to the
1971).
Humans
way one would
also
rotate actual physical objects (Shepard
lifetime of experience of physically manipulating objects in the world.
that these characteristics of
CAD
human
system design. In particular,
and
tactile
We
believe
perception and action should be manifest in
we
believe that there should be an isomorphism
between the physical manipulation of the part
the spatial
and Metzler
have well-developed hand-eye coordination because of a
as displayed
manipulations of a user's hands.
by the
We
CAD
call this
system and
isomorphism
a
10
natural mapping.
would be
A
space.
to
For example, a natural
move
natural
mapping
mapping
for elongating a
for rotating a displayed object
through a corresponding rotation
a hand-held input device
in
displayed bar would be to stretch a hand-
held object. Drilling a hole in a displayed part would be achieved by moving a
hand-held
much
bit
through the desired
and remember (Norman
easier to learn
cognitive burdens imposed by
4.
Controls structured in this fashion are
trajectory.
some
existing
1988),
CAD
and thus
from the
free the user
systems.
The Prototype System
We
CAD
built a prototype
system
and bent sheet metal part design
to test
This choice was motivated by three factors:
ubiquitous part process technology,
2)
We
our key propositions.
domain
as a
(an example part
1)
sheet metal
is
is
chose blanked
shown
metal part geometry
simple enough that
is
system (we
In this section
will refer to
it
we
we
1).
an important and
sheet metal parts exhibit
some
of the
important problems associated with three dimensional part design, and
ideas relatively easily.
in Figure
3)
sheet
expected to be able to implement our
describe our prototype sheet metal
CAD
as the system).
o
o
9J
Figure
1:
The major idea behind
through spatial and
Example blanked and bent sheet metal part
the system
tactile
is
invoke production-like design operations
manipulations of a tool-like input device.
the user literally bends, stretches,
and moves the input device
corresponding real-time modification
screen.
to
to the sheet
in
In our system,
space and sees a
metal part displayed on the
For example, to create a bend, the user positions an image of the input
device on the image of the part by moving the physical tool in space (much like a
three dimensional cursor)
and then bends the hand-held input device while
a
bend
I
is
!
displayed in the image of the part on the screen. (We will
part the virtual part
4.1
and the image of the input device the
image
of the
System Description
The system has two major components:
device, as
designed.
shown
several
2.
body
section,
momentary switches
functions.
may
in Figure
a display screen
and
a tool-like input
The display screen shows an image of the part to be
The input device (Figure
section, the right
3)
has three main components:
the
left
body
and the hinge assembly. Each body section contains
that enable users to select
The hinge assembly has two degrees
be either rotated or translated relative
Figure
The
call the
virtual input device.)
2:
the
two body
sections
one another.
The Prototype System.
spatial position of the input device
Polhemus 3Space sensor mounted
to
and perform desired design
of freedom:
is
relayed to the system software by a
in the right
body
half.
This sensor measures the
absolute position and orientation of the device in three dimensions.
This
12
information
is
map
required by the design functions to
the virtual icon to the actual
input device.
Figure
A
3:
The Input Device
small virtual input device icon, similar in shape to the actual input device,
shown on
the virtual sheet metal piece.
which correspond
to the
The icon
two halves of the
tool.
is
composed
This icon
may
of
be
two
is
rectangles,
moved about
the
planes of the virtual material by the user. Operations are performed on the part at
the position
and orientation
The system allows engineers
of this virtual icon.
to create parts
by performing various production- like
operations on the virtual piece of sheet metal using the
tool.
inputs received from the device are interpreted as function
system's software.
•
The functions
Tactile
and
spatial
commands by
the
available to the user are:
Bend: Pressing and holding the bend button on the input device and bending
the
two halves of the input device performs a bend
position and orientation of the virtual icon.
a
change
The bend function
in the angle of the input device causes
design representation. As the metal
is
bent,
in the sheet metal at the
is
interactive;
an associated change
in the
two physical constraints are
in the sheet of
•
bend radius
the given
ensured:
metal
is
Pulling apart the
Stretch:
added along the
axis
implemented and the amount
is
two body halves causes new sheet material
determined by the virtual
icon.
same
direction.
simple stretch of metal.
•
number
of edges
and
The resulting change provides more than
a
plane of the sheet and does not change
in the
vertices can be accomplished.
the virtual icon axis in a
manner
function.
Both stretch and shrink violate a
appear
be very useful.
to
all
and moves those
Shrink: Pushing the two body halves together causes sheet material
removed along
be
Using the Stretch and Shrink functions, practically
any deformation which occurs
the
to
This function finds
points of the sheet metal on one side of the virtual icon axis
points in the
of material
conserved.
strict
to
be
similar to the stretch
production metaphor, but
Both stretch and shrink are natural and intuitive
operations on a sheet of metal.
•
Cut: Pushing the cut button causes the icon to change to a red line anchored
at the initial position of the icon.
space, the cut line
sheet metal.
The cut has
•
Punch
menu
•
a
moves around
As
moves
the input device in
When
the user releases the cut button, the cut
width
that
is
punch
is
implemented.
preset by the user.
(currently unimplemented):
of
the user
the screen, but remains anchored to the
Pushing the punch button
outlines to appear on the screen.
outline
and places the center point
can be
moved
The user
in the desired place.
will cause a
selects the desired
The punched hole
interactively after initial placement.
Move: Pressing and holding
the
move button
allows users to
move
virtual icon along the planes of the virtual sheet metal design.
and rotations
in space are
mapped
to
movements and
the
Movements
rotations of the icon on
the virtual sheet.
•
Lock: Pressing the lock button "locks" the virtual sheet to the icon, so that
rotation of the input device causes rotation of the virtual sheet.
This feature
14
is
selected often
by users
understand the three-dimensional shape of
to help
the part being designed.
Undo: Pressing
•
to the state
The system has been
many
undo button
the
a vehicle for exploring
other features and functions that
be used
to
restores the
geometry of the
our central research ideas, but there are
would be required before
design parts in industrial practice. In particular,
implemented any way
part dimensions, 3)
to 1)
make curved
cuts, 2)
add dimensional information
be incorporated into our system
if it is
to
the system could
we have
not
impose parametric values on the
to
geometry, or
dimensional shop drawings from three-dimensional models.
5.
virtual design
preceding the most recently implemented function.
4)
produce two-
These features should
be useful in an industrial setting.
Discussion
In this section
we
We
prototype.
highlight
first
what we believe are
discuss
what was learned from the construction
of the
how
We
efforts like these
can be evaluated.
the key features of the system. Finally,
we propose
then present
a
few
directions for improvement.
5.1
Evaluating Prototype Systems
To evaluate
the system,
we demonstrated
it
to engineers,
and researchers. Benchmarking against commercial
as the
system lacks
full
CAD
functionality.
CAD
engineering managers,
systems
We
We
good
CAD
systems
will use the information gained in the demonstrations to
to
as
The demonstrations
early test of the system concept.
plan to benchmark the system against commercial
and
inappropriate
The demonstrations were valuable
they elicited insightful and sometimes surprising comments.
thus provided a
is
in the future.
enhance the system
develop hypotheses for benchmarking studies. Some of the key features of
the system
and some ideas
for
system enhancements that arose from the
demonstrations are presented next.
5.2
Key
Features of the System
The two major advantages
of use
of the system over existing alternative concepts are ease
and part producibility. The system
is
easy to use. Although
we have
not
conducted any controlled experiments,
we have observed
that with
one or two
minutes of explanation, a novice user can create a producible sheet metal part
We
design with the system.
attribute ease of use to the
design of the input device. The system metaphor
punch, and cut are directly analogous
is
system metaphor and
of a production process.
production operations for
to the
metal cutting combined with manual press brake bending.
not have
common
Stretch
CNC
to the
Bend,
sheet
and shrink do
analogies for blanked and bent sheet metal parts, but they are
This metaphor makes remembering the system
intuitive physical operations.
commands extremely simple and
contributes to the ease of production of the
resulting parts.
The design
of the input device also
few operations
natural
that can be
hand motions and
bending the
in
part,
device on the screen.
makes
the system easy to use.
invoked by the user and they are
visible button actions.
moving
the
hands
results in
all
There are only
combination of
a
Bending the input device
movement
to
Holding the cut button and moving the hands results
modify
a real part
The undo feature further enhances
The second major advantage
and therefore are easy
results
of the virtual input
cutting of the part. All of these actions are directly analogous to the
might be used
a
way
remember and
to
in a
a real tool
invoke.
usability.
of the system, part producibility,
is
ensured by both the
production metaphor and by the production constraints. The production metaphor
for
design constrains the user
possible combination of
There
to create physically realizable parts.
commands
is
no
that will result in a design with a face that
is
not
closed or a line dangling in space. Production constraints also help to ensure that
not only will the design be physically realizable, but that
it
can be
made by
the sheet
metal process. The system imposes production constraints primarily in the bend
When bending, the radius of the bend is specified by a program
When the bend radius parameter is properly set, the user can not create
bend radius that is not realizable by the bending process. An additional constraint
function.
parameter.
a
imposed by the system
The goal
is
that
bends can not be made across an existing bend.
of imposing part producibility
functions.
With these functions,
a flat sheet, e.g. a tab could be
rest of the part.
We
it is
was
violated
by the
stretch
and shrink
possible to create a part that will not unfold to
extended such that
if
unfolded
decided to include stretch and shrink to
it
would overlap
facilitate
the
modification
16
of part dimensions at the expense of this potential problem.
by automatically monitoring
can no longer be unfolded
5.3
This could be addressed
condition and alerting the designer
this
when
the part
to a flat sheet.
Suggestions for Improvement
There are a number of ways
sheet-metal-specific
applies to
•
many
We
which the system can be improved.
in
improvements
will discuss
here, but believe that the substance of the ideas
design domains.
Intermediate part geometry: Sheet metal parts are formed by blanking (cutting)
then bending.
would
Being able
facilitate the
view part geometry
at
perform operations on the unbent sheet metal part
to
design
activity.
In effect this
an intermediate point
would allow
the designer to
production process.
in its
Operations performed on either the bent or unbent sheet would cause changes
in
both sheets. This feature, for any
CAD
system,
would allow
the user to
evaluate constraints which apply to parts at different points of the production
process.
•
Performance evaluation:
would help
the engineer
between scrap
like stress
•
costs
A
model
cost
make
and bending
and weight, would
of the sheet metal production process
trade-offs
costs.
between cutting and bending, or
Other types of performance evaluation,
also be useful.
Production constraints: The system, as currently implemented, has some
default values for design operation parameters.
bend radius
is
For example, a
Similarly, a default cut
width has also been defined. Other default
characteristics for other functions could be defined
system. For example,
require a
set
minimum
and incorporated
some manufacturing processes
distance between bends.
and violations flagged. The
within design constraints and
is
result
is
and perpendicularity on
cuts
This
for sheet metal
minimum
into the
bending
distance could be
that the user automatically operates
made aware
Another type of producibility constraint
parallelism
minimum
defined and can be changed for different types of sheet metal.
that
of violations as they occur.
might be useful
is
to
impose
and bends. This would make parts
17
easier to produce.
feature might be
•
Automation:
Since this constraint must often be violated in practice, this
implemented
may
It
and use the system
amenable
be possible
automate some aspects of sheet metal design,
to
interface only for non-routine or other types of design not
automation. For example,
to
some types
as a user option.
shape of the interfaces of the bracket
(such as other parts, tubes, cables,
The system could automate
it is
to other parts, as well as the other objects
etc.) that
the design of
brackets in a given design, and leave the
6.
possible to automate the design of
by specifying the position, orientation, and
of sheet metal brackets
the bracket
many
more
must not
intersect with.
of the simpler sheet metal
difficult brackets to the engineer.
Summary
Engineering design
is
complex work involving the juggling of
information, such as customer needs, process limitations,
a great deal of
and material
constraints.
This can lead to a situation in which engineers are overwhelmed by design
complexity and do not catch and correct design flaws. Economically punishing and
even
We
tragic failures can result.
have argued
that the complexity of design can be
design knowledge into the functionality of the
CAD
complexity of
CAD
reduced by incorporating more
system and by reducing the
system operation. These goals can be achieved by utilizing a
production-like design metaphor, by providing design evaluation, by
well understood constraints, by automating detailed design tasks,
incorporating natural mappings into the
We
prototype to
built a research
part producibility.
enhancements
Much
input device combine to
that, in
an industrial
make
setting,
Demonstrations of the prototype provided ideas for
to the
could be done
It
We believe
that the production
placed on the bend operation can contribute to enhanced
system.
to further
ease the cognitive burden on the engineer.
research and system development
presented.
We learned
spatial-tactile
the system very easy to learn and use.
we
system controls and display.
test these ideas.
metaphor and the ergonomics of our
constraints like those
CAD
embedding
and by
would
is
needed
to refine the
design goals
More
we have
also be valuable to apply these design goals in different contexts.
18
We
believe that the resulting
CAD
systems will support the engineer more
effectively, allowing a focus of attention
on
critical or creative
design tasks.
19
Bibliography
The Challenge of CAD/CAM."
Managing Complexity in High Technology
Adler, P.A. "Managing High Tech Processes:
Glinow and Mohrman
Organizations,
New
(ed.)
in
Von
York: Oxford University Press, 1990.
"The Mental Rotation and Perceived Realism
J. Sanford and J. Foley.
Computer-Generated Three-Dimensional Images." International Journal of ManMachine Studies, Vol. 29 (1988): pp. 669-684.
Barfield, W.,
of
Bemis, S.V., J.L. Leeds, and E.A. Winer. "Operator Performance as a Function of
Type of Display. Conventional Versus Perspective." Human Factors, Vol. 30, No.
2
(April 1988): pp. 163-169.
Bull, R.J.
"Concept to Prototype Using Computing Techniques," Proceedings of the
Mechanical Engineers Conference on Effective CADCAM '87/Towards
Institute for
November
Integration,
1987.
and J. Ishimaru. "Computer Graphics and Animation Come to Ship
Designing," Computers in Mechanical Engineering, Vol. 3, No. 1 (July 1984): pp. 33-
Calkins, D.E.
42.
Computer Aided Design
Report. Vol. 10, No. 5
Computer Graphics World. "A Roundtable on
(May
1990).
Solid Modeling" June, 1988:
pp. 41-
51.
Crombez, G.R. "Chrysler Builds
No. 3 (March 1988): pp. 44-51.
a Better
Network." Manufacturing Systems, Vol.
6,
Cutkosky, M.R. and J.M. Tenenbaum "CAD /CAM Integration Through Concurrent
Process and Product Design," Proceedings of the ASME Winter Annual Meeting
Symposium on Intelligent Integrated Manufacturing Analysis and Synthesis, 1987:
pp. 1-10.
DeMatthew,
No.
1
T. "Productivity
(Jan. 1989):
Dutton,
B.
No. 7 (July
by Computer-Aided Design." Automation,
Vol. 36,
pp. 52-54.
"More Tales From the Factory
Floor."
Manufacturing Systems, Vol.
4,
1986): pp. 38-42.
Eade, R. "Screw Machining: An Industry with Nine Lives." Manufacturing
Engineering, Vol. 100, No. 6 (June 1988): pp. 73-76.
"CADCAM-
Institute for
Productive or Counterproductive?" Proceedings of the
Mechanical Engineers Conference on Effective CADCAM '87/Towards
Integration,
November
Farrar, D.J.
1987.
"
20
"Design of a Three-Dimensional Tactile Input Device for
Mechanical CAD." Master of Science Thesis, Massachusetts Institute of Technology
Department of Mechanical Engineering, May 1990.
Filerman, M.C.
IEEE Spectrum, Vol.
Fitzgerald, K. "Compressing the Design Cycle,"
1987): pp. 39-42.
24,
Frangini, M. "Racing Team Relies on CAD /CAM to Speed Up Design
Computing Canada, Vol. 16, No. 17 (August 16, 1990): pg. 10.
and
Krouse,
J.
Benefits
Outweigh Costs?" Machine Design,
R. Mills, B. Beckert,
C. Potter.
No. 10
(Oct.
Process,
"CAD/CAM
Vol. 61,
Planning Guide: Do
No. 15 (July 20, 1989): pp. S6-
S12.
"CADCAM
Lansiaux, P.
for
in Rolls-Royce
Aero Engines," Proceedings of the
CADCAM '87/Towards
Institute
Mechanical Engineers Conference on Effective
Integration,
November
1987.
Machover, C. and R.E. Blauth. The
Bedford,
MA:
CAD/CAM
Handbook. Computervision,
1980.
Majchrzak, A. and H. Salzman. "Social and Organizational Dimensions of
Computer-Aided Design." IEEE Transactions on Engineering Management,
EM-36, No. 3 (Aug 1989): pp. 174-179.
Manji,
J.F.
"Time
to
Market
is
Critical."
Automation, Vol.
36,
No. 7
Vol.
(July 1989):
pp.
36-41.
Mantyla, M.
An
Introduction to Solid Modeling.
Rockville, Maryland:
Computer
Science Press, 1988.
McKim, R.H.
Miller, G.
Experiences in Visual Thinking.
NASA
Number
"The Magical
Bulletin, Vol. 63 (1956):
Publishers, 1980.
Seven, Plus or Minus Two,"
Psychological
Technical Briefs, August 1989.
of
Everyday Things.
Pentland, A.P. "Towards an Ideal 3-D
Man Machine
Park,
PWS
pp. 81-97.
Norman, D.A. The Psychology
the
Boston:
CA. SPIE
Petroski, H.
CAD
New
York:
Basic Books, 1988.
System," in Image Understanding and
Center, SRI International, Menlo
Interface, Artificial Intelligence
Vol. 758, 1987.
To Engineer
is
Human,
New
York,
New
York:
St.
Martin's Press,
1985.
21
"Solid Modeling and the Interface Between Design and Manufacture,"
IEEE Computer Graphics and Applications, Vol. CGA-4, No. 7 (July, 1984): pp. 52-59.
Pratt, M.J.
Roberts, A. "3-Draw:
A
Three-Dimensional Computer Aided Design Tool," Master
Technology Department of Mechanical
of Science Thesis, Massachusetts Institute of
Engineering, December, 1989.
Salzman, H. "Computer-Aided Design: Limitations in Automating Design and
Drafting." IEEE Transactions on Engineering Management, Vol. EM-36, No. 4 (Nov
1989): pp. 252-261.
Schmandt, CM. "Spatial Input/Display Correspondence in a Stereoscopic Computer
Graphic Work Station." ACM Computer Graphics, Vol. 17, No. 3 (July, 1983).
Shepard, R.N. and C. Feng. "A Chronometric Study of Mental Paper Folding."
Cognitive Psychology, Vol. 3 (1972): pp. 228-243.
Shepard, R.N. and J. Metzler. "Mental Rotation of Three-Dimensional Objects."
Science, Vol. 171 (1971): pp. 701-703.
Simon, H.A.
The Sciences of the
Artificial
(2nd Edition).
Cambridge, MA:
The MIT
Press, 1981.
Smith, D.
"CAD/CAM
Productivity
Time
Ratios for Various
Drawing Types,"
Society of Manufacturing Engineers, Dearborn, MI, Mar, 1982.
Teresko,
No. 6
J.
"Speeding the Product Development Cycle." Industry Week, Vol. 237,
(July 18, 1988): pp. 40-42.
Teresko,
J.
"A Report Card on
CAD/CAM."
Industry Week, Vol. 239, No. 6 (March
19,1990): pp. 44-50.
Vasilash, G.S. "Sheet Metal Processing:
100,
No. 4 (April
Sophisticated Simplicity"
Production, Vol.
1988): pp. 53-58.
Velocci, T. and J. Childs "French Aerospace: Global Leadership." Aviation Week
and Space Technology, Vol. 133, No. 9 (Aug 27, 1990): pp. S5-S24.
Conway, "Computer Applications in
Annual Review of Computer Science, Vol. 3 (1988): pp. 349-387.
Voelcker, H.B., A.A.G. Requicha, and R.W.
Manufacturing,"
Waldern, J.D., A. Humrich, and L. Cochrane "Studying Depth Cues in a ThreeDimensional Computer Graphics Workstation." International Tournal of ManMachine Studies, Vol 24 (1986): pp. 645-657.
22
A ppendix
System Description and Specifications
Figure A:
The system
is
shown
in
Figure A.
of the system are described.
The Sheet Metal
CAD
System
Below, the software and hardware specifications
23
Software
The system software
SunPHIGS
version
is
written in the
C programming language
The system represents sheet metal using
Sheet metal data
1988).
faces
composed
using
calls to
1.1
is
of linked
a full
winged-edge data structure (Mantyla
represented as vertices, edges connecting vertices, and
of edges.
lists
Hardware
The software runs on
computer
a
Sun 4-330 SPARC workstation. This
equipped with
is
a
CXP hardware
32-bit
RISC -based
graphics accelerator using an 8-bit frame
buffer.
A
Polhemus 3Space
device.
Isotrak Sensor
System 4
is
embedded
in the
hand-held input
This sensor transmits three-dimensional orientation and rotation
information to the Sun computer's RS-232C serial port
at
19200 baud.
Inputs from the push-buttons and the bend sensor potentiometer on the input
device are delivered to the computer through a Data Translation 1414 Interface
Board. This
VMEBus
The input device
board
is
equipped with
relays both analog
and
a 12-bit analog-to-digital converter.
digital data to the
board.
4
Polhemus,
Inc.,
Colchester VT.
MOO
!G5
computer through
this
Date Due
jMt \i&*
Lib-26-67
IIT
3
TOflO
LIBRARIES DUPL
0D7D1S3D
S