Rapid Tooling Technologies & Industrial Applications
Rapid Tooling Technologies & Industrial Applications
Rapid Tooling Technologies & Industrial Applications
Technologies and
Industrial Applications
edited by
Peter D. Hilton
Technology Strategies Group
Concord, Massachusetts
Paul F. Jacobs
Laser FareAdvanced Technology Group
Warwick, Rhode Island
TM
ISBN: 0-8247-8788-9
This book is printed on acid-free paper.
Headquarters
Marcel Dekker, Inc.
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540
Eastern Hemisphere Distribution
Marcel Dekker AG
Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-261-8482; fax: 41-61-261-8896
World Wide Web
http:/ /www.dekker.com
The publisher offers discounts on this book when ordered in bulk quantities. For more
information, write to Special Sales/Professional Marketing at the headquarters address
above.
Copyright 2000 by Marcel Dekker, Inc.
Neither this book nor any part may be reproduced or transmitted in any form or by
any means, electronic or mechanical, including photocopying, microlming, and recording, or by any information storage and retrieval system, without permission in
writing from the publisher.
Current printing (last digit):
10 9 8 7 6 5 4 3 2 1
PRINTED IN THE UNITED STATES OF AMERICA
Preface
Rapid Tooling: Technologies and Industrial Applications describes the current, albeit quickly evolving, state of rapid manufacturing (RM) and rapid
tooling (RT), and identies the basic aspects of each commercially available
RP&M system. The primary goal of this book is to provide useful information
to individuals and organizations considering the use of rapid tooling technologies in product development. It discusses the benets of using rapid prototyping and manufacturing (RP&M) technologies in the development process, and
identies complementary technologiessuch as computer-aided design
(CAD), computer-aided engineering (CAE), and computer-aided manufacturing (CAM)that need to be applied in conjunction with RP&M to achieve
maximum benets.
The book is written for people who need to determine whether, or when,
to introduce RP&M into their organization. Engineering managers responsible
for product development or manufacturing processes should nd this book
extremely valuable in providing a background for the use of RP&M within
their organization. R&D managers with product development responsibility
will nd information regarding advanced techniques that their departments
will want to assess, and quite possibly introduce and support. Mechanical engineers, material scientists, and manufacturing/industrial engineers who may be
called on to use RP&M technologies should nd specic information within
this book that is directly relevant to their work. Finally, RP&M technology
and business participants will want to read this book to learn more about the
state of the technology, some of its unique applications, and the likely direction
of its future development.
The RP&M industry has shown various signs of maturation. There has
already been industry consolidation. Several rms have developed essentially
stable market positions, one system supplier has failed, and several others are
losing ground. Rapid prototyping services are available from over 350 service
bureaus worldwide. The competition for business among these rms has
driven prices downward, and reduced prot margins. Somewhat ironically,
the low prices of RP&M parts that have adversely impacted some service
bureaus have caused other organizations to use service bureaus rather than
purchasing, installing, and operating their own RP&M equipment internally.
This, in turn, has adversely affected the sale of equipment from the RP&M
system manufacturers. As a consequence, revenue growth has slowed, or even
reversed, and losses have often replaced prots in the quarterly reports of these
rms.
Notwithstanding all these difculties, the picture is hardly as bleak as
one might initially surmise. At its core, RP&M is really about catching errors
early in the design process, designing better products, reducing product cost,
and getting products to market faster. All four of these benets are surely
coveted by nearly all industries. We believe that the rst order of business
for the RP&M industry is educationwhich is one of the primary reasons
for writing this book. As more companies learn about the time and cost savings
that are possible with RP&M, business opportunities for our industry will
expand. Although awareness of RP&M is certainly much greater than it was
a decade ago, the percentage of those companies that manufacture a physical
product while utilizing RP&M remains pitifully small. To get a sense of this,
the next time you attend a dinner party or a baseball game and the person
next to you is an engineer, scientist, or business manager, ask that person if
they have ever heard of RP&M.
The old saw nothing succeeds like success is truly relevant. Those
organizations that have experienced signicant time or cost savings, or improved product quality through RP&M, become true believers. These rms
continue to use the process over and over again. What does it take to convince
someone who has not achieved these benets that they are real? How does
one show someone that these benets can be applied to his or her specic
application? Is the lack of adoption related to fear of failure? Are these people
afraid that if they recommend the use of RP&M during the kickoff meeting
for their next product development, others will look at them as if they were
from Mars? Perhaps if they read the story of Project Widget in Chapter 3
they will realize what they might be missing if they do not utilize RP&M.
ing, bridge tooling, cast tooling, and production tooling. Chapters 9 through
11 focus on specic applications of RP&M that are currently being employed
in the automotive, medical device, and investment casting industries. The nal
chapter, by Peter Hilton, provides a perspective on the future of RP&M, addressing its likely market penetration and technology growth.
Peter Hilton would like to dedicate his efforts in the publication of this
book to his wife, Joannie Hilton. Paul Jacobs would like to dedicate his efforts
in the publication of this book to his parents, Margaret Veronica Jacobs (1910
1999) and Bertram Lawrence Jacobs (18991975).
ACKNOWLEDGMENTS
Paul Jacobs would like to acknowledge the following individuals whose teaching, counsel, wisdom, effort, support, and vision have helped shape a career:
Gardner Ketchum, Raymond Eisenstadt, and Carl Niemeyer at
Union College
Jerry Grey, Martin Summereld, and Lyman Spitzer at Princeton
University
Gordon Cann, Rolf Buhler, Ken Gustafson, and Bill Hug at Xerox
Corporation
Chuck Hull, Hop Nguyen, Rich Leyden, and Jouni Partanen at 3D
Systems
Terry Feeley, Kip Brockmyre, and Tom McDonald at Laser Fare.
Peter D. Hilton
Paul F. Jacobs
Contents
Preface
Contributors
1.
Introduction
Peter D. Hilton
2.
Process Modeling
Georges Salloum
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Contributors
Anthony T. Anderson
Peter D. Hilton
Georges Salloum Integrated Manufacturing Technologies Institute, National Research Council of Canada, London, Ontario, Canada
Sean Wise
1
Introduction
Peter D. Hilton
Technology Strategies Group
Concord, Massachusetts
This book focuses on the manufacturing portion of the broader rapid prototyping and manufacturing (RP&M) eld. Our interest is in the rapid production
development of relatively low-volume functional parts: parts made out of the
production materials and produced by the production processes. Examples include investment-cast, nickel alloy aerospace engine components and injection-molded polymeric parts (e.g., electronic enclosures). Developing the ability to produce these parts requires developing forming molds or tools for the
parts. Traditionally, the development of such molds or tools is by machining
and heat treating; it requires substantial calendar time and has signicant associated costs. Further, changes to the molds and tools also require signicant
time and costs. Therefore, it is of interest during product development to be
able to quickly produce some rst real parts and to be able to modify the
subsequent parts rapidly based on ndings associated with these rst parts.
We call the ability to rapidly develop molds or tools for moderate volume
parts or products rapid manufacturing.
some subsequent processes. For example, an RP model of the part sought can
be produced and subsequently used as a sacricial pattern to investment cast
the part. Alternatively, a mold can be designed and the patterns for making
the mold can be produced in plastic or wax using an RP technology. These
RP pieces can be used sacricially in the investment-casting process to form
mold inserts in metal.
The various rapid manufacturing processes (to be discussed in this book)
compete against computer numerically controlled (CNC) machining. CNC is
the more mature technology that is threatened by the newer RM technologies.
As is not untypical in these situations, advances are being made in CNC and
related technologies in response to the threat.
It is interesting to postulate how long it will take for RM processes to
replace traditional toolmaking processes. Material presented in this book will
show that RM processes are still under development. It is premature to select
the winning technologies, although some of the losers are already becoming
obvious. The most recent work indicates that we are able to produce molds
for high-volume production using some of the alternative technologies. The
technologies require further eld verication to develop the needed condence
in their long-term performance. However, they are able to contribute to addressing the critical competitive factors of time and quality through reducing
product-development time, improving productivity, and enabling product dimensional control quality.
The benets of a new technology are always weighted against the risks.
Leading users are those who are willing to implement the technologies early,
assuming risks in the hope of achieving competitive advantage. These rms
typically have strong technology competencies and are able to survive startup glitches.
The rate of technology acceptance varies enormously by technology
category and application industry. New electronics technologies that provide competitive benet, particularly software, are generally implemented very rapidly. Materials-related technologies, particularly for transportation applications, typically require decades to achieve substantial market
penetration. Rapid manufacturing technologies have attributes of both software and materials processing technologies. The authors anticipate that the
market penetration of RM beyond the lead users will be quite slow though
steady. Our projection is based on our sense that the mold-making industry is slow to embrace change and that they will need to be pushed by
their OEM (original equipment manufacturers) customers to implement new
technologies.
Introduction
Hilton
ment while minimizing error introduction. Figure 2 lists the computer tools
used to support product development and their roles during the various phases
of the development process. The product and its components are designed on
a CAD system. The CAD models are transferred to a CAE environment for
analyses of product functional performance and of the manufacturing processes for producing the product. The CAD information is also transferred to
those responsible for manufacturing process development and they use it to
design tooling and to create the CAM les for machining operations. The
CAD le is transferred once more to those who will produce prototypes and
patterns using RP. The participants in these various processes may be internal
to the company developing the product or they may be external suppliers of
tools, RP services, or analysis services. Thus, the integration of these engineering software systems to enable direct communication between them as
well as the electronic communication network among the product-development participants provides important leverage in the product development process.
Virtual prototyping is the natural extension of CAE (engineering analysis). It simulates the product functionality and the processes for producing it
prior to development of physical prototypes. Virtual prototyping enables the
design team to perform at least one design iteration without producing hardwarethereby saving time and cost. Virtual prototyping tools also guide in
optimization of the product and the manufacturing process.
IV.
The history and present state of rapid prototyping is reviewed next. The reader
is referred to several textbooks (1,2) for more information on the subject of
rapid prototyping.
that are adhesively bonded to form a layered structure. Cubital uses an ultraviolet-sensitive polymer and a photo-transfer-based approach to create each
complete layer concurrently. Ballastic particle modeling technology directs
polymeric particles at high velocity onto the target, where they adhere to form
a 2D pattern and repeats the process to cause the formation of a three dimensional layered part.
The major commercial technologies are listed in Table 1, followed by
their associated equipment producers, material capabilities, and applications.
Sandia National Laboratories is developing a process to directly produce solid
metal parts. The process, known as Laser Engineered Net Shaping (LENS),
involves the use of a high-power YAG laser that locally melts the metal substrate while metal powder is directed into the resulting melt pool. As with the
previous processes, LENS forms the three-dimensional object by creating a
series of layers.
Each of these rapid-prototyping techniques uses a similar approach to
transfer the part geometry. A solid model of the part is developed on a CAD
system and subsequently sliced by parallel planes to create a series of equalthickness layers. Each of the rapid-prototyping systems recreates these layers
in physical material and builds up the layers to form the part. The distinctions
among the systems are the process for putting down material and the materials
applied.
Recent technology development in rapid prototyping includes both the
renement of existing processes and the development of new processes such
as LENS. Both 3D Systems and DTM have been adding new materials for
their RP processes. 3D Systems with the help of the resin producer, Ciba
Geigy, has been rening resin systems to improve toughness and dimensional
control. They have also added control algorithms (build patterns) to enable the
creation of a honeycomb structure with continuous surfacing. This process,
known as QuickCast, provides advantage when the RP part is used as a casting
pattern. The developments in resin systems in combination with build pattern
renements have enabled 3D Systems to substantially improve the accuracy
capability of stereolithography. In addition, 3D Systems has developed new
models of their rapid-prototyping equipment which have increased forming
speed and enable larger RP parts. DTM has added two new material types to
its RP capabilities; one is a proprietary resin system called TrueForm. Its use
enables more accurate parts and more effective casting patterns. The second
new powder is metal encapsulated in a polymeric coating and the associated
RP process has been named RapidTool. RapidTool, as the name suggests,
represents a process of directly forming metal parts that can be used for tooling
applications.
Table 1
Technology
Equipment manufacturers
Materials capability
Applications
Stereolithography
3D Systems
Epoxy resins
Selective laser
sintering
DTM
Thermoplastics, waxes,
metal powders
Stratasys, Sanders
Cubital
Thermoplastics, waxes
UV curable resins
Introduction
Recently 3D Systems commercialized a second type of rapid-prototyping technology that enables it to produce fast prototypes at lower capital costs.
This prototyping machine called Actua and even more recently ThermoJet
uses multiple heads to extrude thermoplastic resins and form prototypes rapidly. The technology competes with the Z-Corp technology. It does not signicantly overlap stereolithography and therefore represents an extension of
the rapid-prototyping functionality offered by 3D Systems.
Rapid prototyping has appeal and is widely used in industry. However,
the need for physical models may have peaked at some of the technology
application leaders. Computer simulations and virtual prototypes are replacing
some early physical prototypes in the product-development cycle. Whereas
performance simulation and virtual prototyping may become a threat to the
prototyping business, rapid-prototyping technologies are now nding growing
applications beyond prototyping.
10
Hilton
VI.
Introduction
11
This book provides further discussion regarding the use of RP&M patterns
for urethane and metal casting (see Chapters 4, 5, and 11).
An important, emerging application for RP is in the toolmaking (or mold
and die) area. Industry is driven by the goal of reducing the time and cost
of product development while assuring that the product and the process for
manufacturing it are of high quality. More rapid product development means
getting to the market faster, enabling a stronger market position with premium
pricing, and/or improved market share. The importance of product development speed varies among market sectors; in the electronics industry, product
life cycles are short and time-to-market is measured in weeks. For example,
Sony has produced many consecutive models of the Walkman as a means of
keeping up market interest and staying ahead of the competition. The toy industry also has a short product life cycle and a strong need for very rapid
product development. Toys are developed in the spring and summer for all
production and distribution timed to meet the holiday shopping season. The
automotive industry is also competing on the time and cost to bring new products to market and has reduced the product-development cycle from more than
4 years down to 2 years at the leading companies. Medical product development also seeks to reduce time and cost; however, product life cycles are long
and product-development times are restrained by regulatory approval processes.
For many products such as those noted, the time and cost of producing
the production tools is a signicant portion of the overall product-development
time and cost. This is particularly true of products that will be produced in
large volumes by automated processes (consumer electronics, toys, cars, etc.).
For example, molding, casting, or stamping tools typically require several
months to produce and cost tens to hundreds of thousands of dollars. Therefore, the possibility of positively impacting the time and cost of tooling production is appealing.
Figure 4 illustrates product-development time savings achieved by one
company through the use of rapid tooling and other computer-based technologies. Several approaches for producing tooling based on rapid-prototyping
technologies are at various stages of development. The earliest efforts were
based on casting technology. The process mentioned earlier for producing
metal castings can be applied to rapid tooling in the form of casting the tooling.
Separately, rapid prototypes are being used in conjunction with a process
known as Keltool to produce tools quickly. Keltool was developed by 3M and
licensed to Keltool, Inc., which was recently acquired by 3D Systems. The
Keltool process enables the reproduction of a physical part in metal. Within
the context of rapid toolmaking, an RP model of the tool is produced and sent
Figure 4 Time savings with rapid tooling and other computer-based technologies.
*: Includes initial process parameter selection; **: verication only.
Each of these processes has the advantage of lower costs and times when
compared to traditional hard-tooling processes; however, these processes have
limitations as well. The limitations are in two areas: dimensional control and
long-term performance of the resulting tools. Dimensional control limitations
may mean that some postprocess machining will be required, which adds time
and cost. Durability limitations may make some of these tools most appropriate to lower-volume production applications and to prototyping by the manufacturing process, for higher-volume applications.
Rapid-prototyping technologies are of interest to the automotive industry because of their ability to create early part prototypes to visualize design
concepts directly and for their contribution to prototype tooling to enable more
substantial prototypes later in the process for testing and evaluation, as well
as to test the nal manufacturing processes (see Chapter 9). In the medical
products industry, manufacturing of orthopedic implants, rapid prototyping is
used for visualization and to check out dimensional ts, and as a means of
low-volume production of cast components (Chapter 10). The aerospace engine industry also uses investment-casting technology to produce low-volume
complex parts. Again, rapid prototyping provides advantage in prototyping
casting processes (Chapter 11).
REFERENCES
1.
2.
PF Jacobs. Rapid Prototyping and Manufacturing: Fundamentals of Stereolithography. Dearborn MI: SME, 1992.
PF Jacobs. Stereolithography and Other RP&M Technologies. Dearborn, MI:
SME, 1996.
2
Process Modeling
Georges Salloum
National Research Council of Canada
London, Ontario, Canada
I. INTRODUCTION
Three facets of product and process optimization involve the simulation of
structural behavior, material ow, and solidication. Simulation technology
is used to improve part and mold design and for the optimization of die casting,
injection molding, blow molding, and thermoforming operations. Such net
shape material processing techniques, molding, forming, and casting are vital
to the mass production of single or integrated components for numerous industry sectors such as automotive, packaging, appliances, electronics, telecommunications, medical, leisure, and sports. The success of these industries in responding to rapidly changing customer demands will depend on the ability
to develop and apply state-of-the-art technology in collaboration with other
partners. Instead of the traditional method, where product and manufacturing
engineering follow in a sequential order, the emphasis is put on developing
the product and its production process concurrently.
As a result of market globalization, the appearance of high-performance
materials, increased product complexity, and geographical variation of raw
material and labor costs, more pressure is put on the material processing industries and end-product manufacturers for constant innovation and process optimization. Their customers demand for higher quality standards and lower
costs often presents a considerable challenge beyond the reach of the individual companies. Despite the introduction of tools like computers, advanced
II.
molds, parting line molds, hot runner molds, multicavity molds, cam action
and multislide molds, lost core molds, and collapsible core molds are examples
of the potential degrees of freedom available to the designers and mold makers.
There are several types of processing machinery to be considered during the
product-process development stage such as vertical and horixontal presses
with or without rotary and shuttle tables.
Concerning the choice of the material for the product itself, there are
a great number of polymers, metal alloys, and composites to be evaluated.
Commodity, engineering, and high-performance materials are available in specic or customized compositions for a given application or market. A material
demand prole should be prepared taking into consideration the structural,
functional, environmental, processing, and recycling requirements. Threedimensional computer-aided design/computer-aided manufacturing (CAD/
CAM) systems operating on workstations and personal computers are available to assist the designers and process engineers. They can access various
commercial computerized databases provided by the material suppliers for the
selection of the most suitable material.
The designer can apply computer simulation to conduct structural analysis, evaluate part functionality, and verify that the part stands up to the demands that will be encountered under service conditions. Beyond this, experienced designers follow accepted practices which directly inuence the weight
and strength of parts. They know it is important to avoid sharp corners and
maintain relatively constant wall thickness to minimize stress concentrations.
As the part design evolves, it is also important to evaluate how various
design options or alternatives inuence how the part can be manufactured.
Some basic ground rules with respect to the number, location, and size of
gates and draft angles should be used to facilitate the lling and the ejection
of the part from the mold. However, to produce parts of high quality, the
product designer or the original equipment manufacturer must cooperate
closely with the mold maker and the material supplier. At the same time, the
quality of a part in terms of strength, appearance, and dimensional tolerances
is directly dependent on how it is produced. It is crucial to recognize that even
a perfectly designed part can be ruined and fail specications if it is formed
under inappropriate conditions.
2.
3.
IV.
Injection molding and die casting are the most important processes for the
production of three-dimensional parts from plastics, metal alloys, elastomers,
composites, and, increasingly, metal and ceramic powder. These processes
involve the mixing, melting, and/or heating of a material followed by its injection, under pressure, into a mold where it subsequently solidies. The injection
step involves the delivery of the melt from the tip of the cylinder or screw
through the sprue, runners, and gate into the cavity. During the lling stage,
the material ows into the mold under pressure or by gravity casting.
The lling stage is concerned with the transient nonisothermal ow
(laminar or turbulent) of the uid. When the cavity is lled, a high-packing/
intensication pressure is applied to force additional material into the mold.
The purpose of this step is to introduce sufcient material into the cavity in
order to compensate for shrinkage during the solidication stage and to ensure
proper lling of complex parts. The increase of the material density in the
cavity is responsible for the rapid increase of the pressure during this stage.
Simulation models are useful for the prediction of the lling pattern,
short shots, voids and weld surfaces, pressure, velocity and temperature distributions, and the overall cycle time. Figures 2 and 3 show the predicted lling
pattern during the injection molding and die casting of two complex parts. In
the solidication stage, a continuous decrease in cavity pressure is observed.
22
V.
Salloum
Process Modeling
23
24
Salloum
Injection and injection stretch blow molding with accurate dimensional control of bottles for the food, medical, and industrial packaging industries.
Vaccum and plug-assisted forming with applications in food and
medical packaging such as formed trays and blister packages
Pressure and twin-sheet thermoforming with applications in panels
and control cabinets, household, and consumer products.
Drape and matched mold forming in automotive and industrial packaging
VI.
CONCLUSIONS
Numerical simulation techniques that account for heat transfer, uid ow,
phase transformations, and stress deformation are generic and benecial to
processor, equipment, and material suppliers and to end-product manufacturers. The only practical means to achieve rapid tooling and the production of
high-quality parts is through the integration of part and mold design with process development. Concurrent process modeling, design analysis, and optimization will improve product performance, prevent or control processing defects, and shorten production time.
The interplay of part design, tool design, material properties, production
conditions, and part quality is extremely complex and involves a matrix of
many variables. It is not reasonable to expect a team of engineers to deal
Process Modeling
25
with these complex interactions and optimize casting and molding processes
without the use of modern nite element methods.
In addition to modeling and simulation techniques, once the rst prototypes have been produced, other performance tests should be conducted under
service conditions. Among the tests most relevant to various applications are
the evaluation of chemical and impact resistance, clarity, substance absorption and degradation, permeability, microcracking, delamination, surface
roughness, discoloration, and so forth.
Computerized process simulation can be used to monitor the inuence
of design alternatives on processability of the part and to select operating
conditions that assure the required part quality. It is clear that nite-element
simulation increases process and product reliability.
REFERENCES
1.
2.
3.
4.
5.
6.
3
Rapid Product Development
Paul F. Jacobs
Laser FareAdvanced Technology Group
Warwick, Rhode Island
I. INTRODUCTION
In 1999 over 99.99% of all injection-molded plastic parts manufactured
throughout the world will be created by tools that were either (a) machined,
(b) formed by electrical discharge machining (EDM), or (c) generated by some
combination of these methods. Production tooling is typically fabricated from
steel, with aluminum used for molding smaller quantities. Machining was formerly done manually, with a toolmaker checking each cut. This process became more automated with the growth and widespread use of computer numerically controlled or CNC machining. Setup time has also been signicantly
reduced through the use of special software capable of generating cutter paths
directly from a computer-aided design (CAD) data le.
Spindle speeds as high as 100,000 rpm provide further advances in highspeed machining. Cutting materials such as cubic boron nitride, which approach the hardness of diamond while possessing outstanding thermal conductivity, have demonstrated phenomenal performance without the use of any
cutting/coolant uid whatsoever. As a result, the process of machining complex cores and cavities has been accelerated.
The good news is that the time it takes to generate a tool is constantly
being reduced. The bad news is that even with all these advances, tooling can
still take a long time and can be extremely expensive. Six months and
$250,000 is not uncommon for a large, highly detailed tool involving numer-
28
Jacobs
ous narrow slits, high-aspect-ratio bosses, shutoffs, and multiple slide actions.
Three months and $30,000 is routine for a small mold with only moderate
complexity.
Many executives now realize how vital it is to move new products to
market rapidly. A company able to launch a quality product ahead of their
competition not only realizes 100% of the market before rival products arrive
but also tends to maintain a dominant position for a few years even after
competitive products have nally been announced (1). For most products,
these two advantages are dramatic. However, when a new version of a laptop
computer has an effective product life of only 18 months, being rst to market
can be critical.
Rapid Product Development is now a key aspect of competitive success.
Figure 1 shows that only 37% of the product mix from the average industrial
or electronics company is less than 5 years old. For companies in the top quartile,
the number increases to 1525%. For world-class rms, it is 6080% (2). The
best companies continuously develop new products. At Hewlett-Packard, over
80% of the prots result from products less than 2 years old! (3).
II.
Let us consider a typical product development cycle for a Widget. One Friday
afternoon in April, a senior engineer named John suddenly gets an idea for a
29
new type of device that he believes people really need, want and will buy. He
then spends a few hours sketching the fundamental characteristics of the idea
while jotting down some notes about potential features and benets. Just after
John thinks this is a great idea, he also wants to get a co-workers opinion.
However, he does not want to look silly. Realizing that there is not enough
detail at this point, John decides to take the idea home and sleep on it. If
it is really such a great idea, it will still be a great idea the next morning!
However, the next morning dawns cold and gray. While showering, John
thinks: How many people will actually buy a Widget? At what price? How
much should a Widget cost to make? How would one make a Widget? What
materials should one use? What process?
When John arrives at work the following Monday, he is both elated and
frustrated; elated that he has the kernel of an idea that could be really signicant, and frustrated that there are so many unanswered questions. He needs
some feedback. So, John talks to Harry. Harry is absolutely convinced that
nobody needs a Widget. Every time John tries to explain how terric a Widget
would be, Harry presents three reasons why it will never work, why it will
cost too much to make, and besides, he heard that some company was already
working on something like that.
Deated, but not totally crushed, John develops the idea for another 2
weeks. He prepares more detailed sketches, thinks about key functions, forms
ballpark estimates of what it might cost to make a Widget, comes up with
an improved version of the idea, and sharpens his arguments in preparation
for further discussions. This time John talks with Nancy. Nancy thinks that
the general idea of a Widget is great, but she has no idea how large the market
might be. She does believe that whatever the market is, it is likely to be price
sensitive. Nancy feels that since people have never had Widgets before, they
clearly are not necessities; rather they fall into the nice-to-have-but-not-essential category.
Furthermore, Nancy thinks that the Widget should be mostly made out
of plastic to keep the price down. Because the intended use is very demanding,
it must also be tough, so something like glass-lled polycarbonate is probably
appropriate. Nancys nal recommendation is that the idea is sufciently interesting that it would be worth having a meeting with key people from Marketing, Product Design, Engineering, Production, and Sales.
George, who is the VP of Marketing, happens to be on vacation. Edward,
the VP of Sales, is at a convention in Boston. The earliest possible time for
the initial meeting is the following Monday. Note that over 3 weeks have
elapsed After Concept Germination or ACG until the rst meeting is held
to even discuss the topic.
30
Jacobs
The meeting, scheduled to start at 9:00 am, actually starts at 9:14 because Bill, from Product Design, was working on a change to another product
that is now behind schedule. Laura and Andrew, from Engineering, need to
leave at 9:55 because they must attend a critical quarterly review at 10:00
sharp. With allowance for coffee and a statement of why everyone is here,
John has only 37 min to describe what a Widget is, how it would basically
work, what are its benets, why people would buy one, and roughly what one
would look like.
George and Ed do not understand the concept drawings, whereas Bill
immediately recognizes that, as designed, the Widget would be nearly impossible to build at a reasonable cost. Richard, from Production, agrees that some
design changes will be needed to simplify the manufacturing process. Laura
thinks the basic idea is good but is concerned about potential thermal problems, and Andrew is already developing variations in his mind. There is not
enough time to establish a consensus, but George agrees that Jennifer from
Marketing would be a good person to look into the potential Widget market.
Bill will develop an initial CAD model from Johns sketches and run them
past Richard. Laura says that she can look into the thermal issues. They all
agree to a second meeting next Monday.
At the second meeting, George explains that Jennifer was working on
a critical project for Division B and, unfortunately, could only assemble very
fragmentary information regarding the potential Widget market. Ed was unexpectedly called to a sales meeting in Denver to establish booking targets for
QIII and could not attend. Laura started a thermal nite-element analysis (TFEA) but realized that she did not have critical dimensions or material property
data and could not proceed without further information. Andrew presented
some concerns regarding excessive deections due to large bending moments.
Bills CAD design had been started but got bogged down when he could not
interpret one aspect of the drawing and three phone calls to John only resulted
in playing telephone tag. Note that 1 month has passed and we now have a
quasi-CAD design, a fragmentary market analysis, and some vague technical
concerns.
During the week Bill meets with John, claries the confusing aspect of
the sketches, and completes a rst-level CAD design, which he forwards to
Richard in time for the third meeting. Because George and Ed may have difculty interpreting the CAD representation, John and Bill decide to send the
CAD le to a local service bureau, ProtoMetrics, to have a full-size model
built by a rapid prototyping and manufacturing (RP&M) system. They are not
sure about cost and also realize that there is no charge number for this task,
as the work on Project Widget has not been approved by Finance and Account-
31
ing. John thinks he can pay for the RP&M model from discretionary funds
but realizes that he better meet with Eric, the VP of Finance and Accounting,
to establish a budget, organize the project team, and assign charge numbers.
John contacts ProtoMetrics and discovers they are currently swamped
with work. The earliest they will be able to deliver the part is 9 days. Consequently, the next meeting is moved to the following Thursday. George, Ed,
Jennifer, John, Bill, Richard, and Andrew can make the new meeting date.
However, Laura is presenting a paper entitled A Finite Element Analysis of
Conformally Cooled Tooling at a conference in Dearborn and cannot attend.
At the fourth meeting, 6 weeks ACG, Jennifer presents data implying
that there could be a signicant market for Widgets. Also, Nancys instinct
that this market is likely to be price sensitive was correct. After speaking with
Richard, Bill realizes that the design will need changes to reduce manufacturing costs. Ed mentions that he spoke with some of his sales team at the Denver
meeting and they seemed excited about the Widget idea. Andrew points out
that Laura had some concerns regarding thermal issues, but she is not here to
present them and he does not feel condent that he can properly represent her
ideas. He also believes that excessive bending moments may lead to distortion
problems, so material properties and section thickness values may be critical.
Everyone passes the RP&M model around the conference table, asks questions, and begins to get a sense of what a Widget looks like, feels like, and
roughly how it would work.
The group agrees they are spending a lot of time on these tasks and that
Project Widget should be formally launched. A need-to-know list is generated
by John. George assigns Jennifer market assessment responsibility. Bill is
completely overloaded on his current task, so he recommends that Donna,
from Product Design, be assigned to the team. She is excellent and has just
successfully completed work on a major project. Laura will work on the thermal analysis when she returns, and Andrew will continue to evaluate critical
deection issues.
John proposes that a regular Project Widget meeting be held each Monday at 9:00 am. Eric assigns Susan, from Accounting, to assist John with
developing a budget. They will also establish a schedule and work breakdown
structure to assure that all key tasks are identied.
The following Monday, the fth meeting takes place (now 7 weeks
ACG). Susan points put that because no approved budget exists for Project
Widget, it will take some time to complete a schedule, personnel loading, work
breakdown structure, and program costing. In the meantime, people should
minimize their involvement on the Widget effort and charge whatever time
they do spend to special account number 99007. After Susans comments,
32
Jacobs
the team decides to pause until Project Widget is formally approved by Conrad, the Division Executive VP, as well as Eric. Concerned that vital momentum is being lost, John sends a memo to management summarizing the work
to date, the initial market estimates, and includes a photograph of the RP&M
model. He forwards copies of the memo to all personnel on the need-to-know
list.
After two additional meetings involving Conrad, Eric, Susan, and John,
Project Widget is nally approved, but with a budget 20% lower than Johns
initial estimate. Erics nal comment, made in the hallway after the second
meeting had concluded, was that John should feel particularly fortunate, as
no other special projects had been approved by Conrad this year.
However, John does not feel particularly fortunate. It is now 9 weeks
ACG and he has this disturbing feeling in the pit of his stomach that somewhere out there, someone else may also be working on their own version of
a Widget. Furthermore, momentum, enthusiasm, and esprit de corps have
all suffered during the 2-week wait for an approved budget. Also, he just
learned that Bill has been reassigned to his former project. Although Donna
may be terric, she is utterly unfamiliar with the current design.
Ten weeks ACG, the sixth Project Widget meeting is held. Susan explains the new budget, schedule, and work breakdown structure. Donna is
introduced to the group and notes that Bill gave her a copy of the current
Widget CAD le as well as the RP&M model. She asks a few questions related
to some of the geometric characteristics, and Andrew explains that they were
required to increase stiffness and reduce deection.
Laura hands out copies of her initial thermal analysis and notes that
excessive heating may indeed occur in two locations. However, until more
detailed T-FEA results are available, based on actual thermal property data
for the proposed material, she cannot be certain about the accuracy of the
predictions. Ultimately, the only way to be condent of the thermal design is
to test a true prototype, injection molded in the intended 30% glass-lled
polycarbonate. Unfortunately, this will require prototype tooling.
Jennifer presents an updated marketing analysis. Her preliminary estimate suggests that the Widget market could reach $80 million this year, $120
million next year, and $180 million in year 3. Into year 4, things become
fuzzy due to potential obsolescence issues and uncertain levels of enhanced
performance in the future. Nonetheless, conservative estimates indicate that
the total market over a 5-year product life cycle could exceed half a billion
dollars.
At this point, everyone in the room is excited. Donna agrees to meet
with John and some local toolmakers to establish estimates of the cost and
33
schedule for prototype tooling. George and Jennifer state that they would love
to have about 200 marketing test samples to generate response from buyers
at major retail outlets. Ed would also like to have about 150200 prototypes
to get some feedback from his salesmen.
During the week, Donna and John meet with three tool and die shops.
They are told that 400 prototypes is a really nasty quantity. If they needed
only a dozen, then soft tooling using an RP&M master, silicone RTV, and
two-part polyurethanes might sufce. Although the mechanical and thermal
properties of various polyurethanes would not be identical to 30% glass-lled
polycarbonate, at least the cost would be low and they could have their parts
within a few weeks.
However, if it is critical that they have true prototypes, injection molded
in glass-lled polycarbonate, then all three toolmakers suggest aluminum prototype tooling. It will cost less than steel tooling and could be ready in 12
weeks. Still, CNC-machined aluminum tooling will be difcult to amortize
over only 400 prototypes. Also, if there are any additional product design
changes, tooling rework can be expensive and will push the delivery date out
even further!
Carefully inspecting the RP&M model, one of the toolmakers notices
a small undercut which would require a slide action. This will further increase
the cost of the prototype tooling and extend its delivery. He inquires if the
design could be changed to eliminate the undercut? Donna says that she will
look into a design modication, will develop a new CAD le, and also have
a second RP&M model made. Donna and John leave the toolmaker realizing
how important it was that this problem was detected now and that an iteration
of the design should not be too difcult because RP&M models can be built
relatively quickly and inexpensively.
By the seventh meeting, Donna has made subtle changes to the CAD
design. However, George, Jennifer, and Ed are not sure what effect these alterations may have on aesthetics. The group decides to purchase four RP&M
models of the new design, one for each toolmaker to improve communication
and reduce bidding uncertainty, and one for the Widget team.
Donna develops a .STL le from her new CAD design, having discovered that this is easier for ProtoMetrics to work with and will reduce their
price as well. She then forwards the .STL le to the service bureau. Unfortunately, they are still swamped with work and can only promise delivery in 7
days. John approves the purchase order, but he must now reschedule the project meeting for Wednesday. The new meeting date is exactly 3 months ACG.
At the eighth Project Widget meeting, Donna passes the new RP&M
models around the room. The undercut has now been eliminated and the aes-
34
Jacobs
thetics look great. Furthermore, the prototype tooling will be simpler, less
expensive, and will be able to be delivered more rapidly. After the meeting,
Donna and John take one new RP&M model to each of the three tool and
die shops, requesting formal quotations on the machined aluminum prototype
tooling.
The following Friday, they have received all three bids. Two of the
shops are quoting 12 weeks and about $50,000. The third shop is quoting 10
weeks and roughly $62,000. Although time is certainly important, it is extremely difcult to convince Susan that 2 weeks is worth $12,000 just for
prototype tooling, so the team decides to go with Central Tool & Dies 12week bid for exactly $50,176.
During the 12 weeks that the prototype tooling is being fabricated, Jennifer starts the layout of the various marketing collateral materials, including
packaging design, photographs, sales brochures, detailed product specications, health and safety compliance information, Underwriters Laboratory
(UL) certication forms, advertising storyboards, and so forth.
Meanwhile, Laura completes a more detailed T-FEA and concludes that
the Widget will probably be operating in a safe regime. However, the temperatures in the two anticipated hot spots remain a concern. Consequently,
Laura strongly recommends that detailed thermal testing of true functional
prototypes, injection molded from the nal intended material, will be required
to establish the actual safety margin, if, indeed, there is a safety margin.
Andrew has also completed a mechanical nite-element analysis (MFEA) and concludes that his original concerns about the parts stiffness were
indeed appropriate. There is an issue with excessive deection causing potential interference during operation. Unfortunately, the margins are sufciently
close that only careful deection measurements on a functional prototype will
truly establish design verication. Also, the potential hot spots identied by
Laura will tend to reduce the modulus of elasticity of the material, which could
further increase the deection, making the problem even worse.
Andrew notes that this is a classic example of an interactive effect,
where normal operation results in mutually dependent thermal and mechanical
loads. Specically, the increased temperatures in the two hot spots locally
weakens the material, leading to increased deection. Simultaneously, the increased deection slightly alters the thermal boundary conditions, which will
change the temperature distribution. The interactive effects may be quite small
or they may prove to be signicant, especially if the design is right on the
edge of passing or failing to meet product specications. It is precisely this
sort of thing that is difcult to predict analytically and is yet another reason
35
why the team will never have warm fuzzy feelings about the design until
reliable test data have been gathered from a true prototype.
Finally, after numerous calls to Central Tool & Die, John is informed
that the tool will only be 3 days late. Apparently, he should be happy about
this. Because Central has some small injection-molding proof presses, they
could run the rst 20 parts on Friday afternoon. John could then pick them
up in time for the regular Project Widget meeting on Monday. The remaining
380 parts could be run the following week, or the tool could be forwarded to
a local injection-molding shop to run the rest of the parts within a day or two.
John agrees to pick up the rst 20 parts on Friday but decides to wait until after
the Monday meeting to select the injection-molding vendor for the remaining
Widget prototypes. While jotting a reminder in his calendar to visit Central
Tool & Die on Friday afternoon, John happens to notice that this will occur
almost exactly 6 months ACG.
At the next meeting, John distributes the injection-molded prototypes.
Everyone is impressed with their overall look and feel, but nal assembly and
functional testing still remain to be accomplished. Laura and Andrew agree
to start testing as soon as possible. Lauras technician, Joan, is out sick with
the u, but John agrees to help Laura assemble and calibrate the required
thermocouples. Andrew has already carefully calibrated six strain gauges in
preparation for mechanical testing and will apply them to a second prototype.
The accumulation of real test data should begin the next day.
Within 2 days, the results of the functional testing are complete, and a
special Project Widget meeting is called. First, Laura presents the results of
the thermal testing. The measured temperatures are within 10F of the T-FEA
predictions and, indeed, there are two hot spots. At 120% of peak anticipated
loading, the temperatures are still within specication, although the data indicate that one is nervously close to the upper allowable limit.
Next, Andrew presents the results of the mechanical testing. Here, the
results are not especially close to the M-FEA predicted values. Indeed, the
largest deections are occurring very near the highest temperature region. Andrew strongly suspects that the elevated temperatures have reduced the modulus of the 30% glass-lled polycarbonate to a point where the stiffness is no
longer sufcient to keep the maximum deection level within specication.
This is precisely the type of interactive effect that Andrew had mentioned previously. There are two possible solutions. The rst is to increase
the glass loading, which will increase part stiffness and thereby reduce the
maximum deection. The second approach is to increase the section thickness,
which would have a similar effect.
36
Jacobs
The good news about increasing the glass loading is that it will not
require a redesigned part geometry. The bad news is that this will demand
higher injection-molding pressure and induce a more rapid erosion of the active tool surface, thereby reducing tool life. Richard states that he has seen
this kind of thing before and that in his experience, increasing the glass loading
is ne up to a point, but he does not think that it will be sufcient in this case.
Conversely, increasing section thickness will almost certainly work, but it
would require a new CAD model, iterated T-FEA and M-FEA analyses, a
modication to the tool, the generation of another set of prototypes, and yet
another round of functional testing, all of which will consume additional time
and money.
John agrees to call Central Tool & Die and ask them to try another 20
shots in the prototype tool, but this time using 40% glass-lled polycarbonate,
which is about as high as they can reasonably go without introducing serious
injection-molding issues. Meanwhile, Donna will modify the CAD design per
Andrews suggestions. Andrew agrees to help Donna with the modied geometry. Using the M-FEA program, he will perform a parametric analysis to
establish how large an increase in section thickness would be required to
achieve a maximum deection within specication. It would be terric if simply increasing glass loading will solve the problem, but everyone agrees that
it would be prudent to have a backup approach as well.
The next day, Central Tool & Die shoots 20 prototypes in 40% glasslled polycarbonate on their proof press. The parts are basically ne, except
for one thin-wall section which looks a bit ragged. Central feels that this can
probably be solved by increasing injection pressure and they will try that tomorrow. Meanwhile, Laura afxes her thermocouples and Andrew his strain
gauges, for a second round of functional testing.
While this testing is underway, Central tries some variations on the injection pressure and determines that indeed a 10% increase seems to solve
the ragged thin wall problem. Unfortunately, just after Central called John
with this good news, Laura and Andrew bring the data from their tests into
Johns ofce. The new material has had a negligible effect on the thermal
results. However, although the maximum deection has been reduced from
the prior test results, it still exceeds specication. This is not good news.
At this point, John calls another special Project Widget meeting. A lively
discussion ensues. Many ideas are presented. Sales wonders what would happen if we only test to 100% of maximum load instead of 120%. Richard
quickly states that a lot of Widgets would be broken by muscular users, the
company would spend a fortune on product guarantees, and the Widget reputation would take a dive. George then asks Andrew how large a change in thick-
37
ness would be required to meet the specications. Andrew nds the latest MFEA results from his parametric analysis and explains that only an 8% increase
in section thickness should be needed.
Recognizing that this approach may be critical to the success of the
entire project, George then inquires of Andrew, What is your condence
level regarding the computational analysis? Andrew states that M-FEA results are generally accurate to within 10%. However, because part stiffness
increases with the cube of the section thickness, if they went from 8% to just
a 10% increase in section thickness, they would almost certainly gain additional safety margin.
Both John and Laura agree that increasing the section thickness by 10%
is probably a good idea. Ed notes that this means that the part volume will
increase slightly, as will the weight and the material cost, and wonders if any
of these might be problems. Richard says that the increase in material cost
will only be pennies per part. Donna says that she can calculate the increase
in weight from the solid CAD model, but she does not expect it to be more
than half an ounce. Jennifer indicates that although excessive weight could
adversely affect the Widget market, having stronger parts that do not break
under hard usage is undoubtedly far more critical to overall product success.
A group discussion follows. George expresses concern about the impact
of a redesign on the Product Release Date or (PRD). Ed asks John how long
he thinks it will take to come up with a revised schedule and a more accurate
PRD. John mentions that the extent of the delay will depend on a series of
events: How long it will take to modify the CAD design, to build a second
iteration in RP&M, to bring the RP&M model to Central Tool & Die for a
quote on reworking the prototype tool, the time for Central to bid, the internal
approval cycle (as tool rework was not in the original budget), the actual time
it will take Central to re-work the tool, shoot another 20 parts, and the time
it will take Laura and Andrew to complete still another round of functional
testing. Nonetheless, John states that this is really their only option, other than
canceling the project. The entire team concludes that except for the deection
problem; (a) the basic design is terric, (b) the potential Widget market is
substantial, (c) they have made considerable progress, and (d) quitting after
all this work and expense would be incredibly wasteful.
Collectively, a decision is made to redesign the Widget. Donna will
update the CAD design, including Andrews latest suggestions for slightly
thicker walls. Richard will talk to Central Tool & Die to get an estimate of
the cost and time needed for tooling rework. John and Susan will also generate
up-dated costs and schedules. When this is complete, John will meet with
Conrad, Eric, and Susan to secure the incremental funding. Richard will then
38
Jacobs
get a rm bid and schedule from Central for the tooling rework. Andrew notes
that because the redesigned prototypes will likely meet all the specications
and Project Widget will require production steel tooling later anyway, perhaps
they should ask Central to prepare a formal bid for this as well. They can
always go out for multiple bids later. John and Richard agree.
Within a few hours, Donna completes the CAD modications. The critical sections in the region near the maximum deection problem are now 10%
thicker. Donna requests that Andrew check the second-iteration CAD model
(i.e., without any undercut, but with 10% thicker sections). Andrew agrees
that the modications look good. He will go through another M-FEA to be
sure that the maximum deections will remain within specication.
Donna uses the solid CAD model of the second iteration to determine
the weight increase relative to the rst iteration. It turns out that her guess
was close; the weight increase will only be 0.382 oz. Although the changes
are small, Donna thinks that if the budget can handle it, they probably should
have another RP&M model made. She notes it would be ironic if they were
heading toward nal production and everyone was touching and looking over
a model of a Widget that was lighter and thinner than what they were actually
going to produce. She calls John with the results and mentions the additional
RP&M model. He agrees that this is a good idea and will add this to the budget
increment. John also thinks to himself that Bill was right, Donna really is
doing a great job.
Andrew locates some data showing the modulus of elasticity of 35%
glass-lled polycarbonate as a function of temperature which looks reliable
(viz. the graph contains error bars, and the test conditions are well dened).
With these data the FEA predictions should be even closer to the test results.
The original design called for 30% glass loading. Further, the team had already
tested both 30% and 40% glass-loaded material from Centrals proof press.
However, Andrew realized that 40% glass loading was probably pushing
the injection-molding pressures a bit. Perhaps 35% glass loading would provide some safety margin without making life too difcult for the production
molding shop. If the M-FEA data looked good, he would recommend that the
nal production material should be 35% glass-lled polycarbonate. Andrew
believes this would be a near-optimum choicebalancing strength, stiffness,
tool erosion, and ease of manufacturing.
Meanwhile, John introduces Richard to Phil, the key person at Central
Tool & Die. Richard suddenly recognizes that Phil is an old friend from college whom he has not seen in years. After John updates everyone on the status
of the project, he feels comfortable that Richard and Phil will deal with the
39
tooling rework issues. John returns to his ofce to meet with Susan about
developing the new schedule and budget.
The next day Andrew completes the M-FEA analysis on the second
design iteration. As he suspected, the combination of 35% glass-lled polycarbonate and a 10% greater section thickness has reduced the maximum deections below the product specication limits, with a nice margin of safety. He
takes the results to Johns ofce just as the senior engineer returns from Central
Tool & Die. John studies the M-FEA results for a few minutes, congratulates
Andrew on a job well done, and agrees that 35% glass loading is probably
close to optimum.
Four days after the meeting with Phil at Central Tool & Die, Richard
receives a formal quote for the tooling rework: $16,240 and 6 weeks. Richard
calls Phil back, thanks him for such prompt quoting, but inquires if there is
any way that the rework could be done faster. Phil informs Richard that Central
has so much work at present that the dominant part of the 6 weeks actually
involves queue time. Simply put, there are numerous rush jobs, and only so
many machines and toolmakers, so each job basically has to wait its turn.
Nonetheless, because they were old friends, Phil will do his best to try pushing
their job ahead a bit whenever possible.
During the next week, John meets with Conrad, Eric, and Susan to go
over the revised budget. Eric is concerned about the additional $16,000 for
the tooling rework. This time, somewhat surprisingly, it is Conrad who points
out that they have already spent over $500,000 on Project Widget, including
burdened labor, RP&M models, prototype aluminum tooling, FEA, and functional testing. The additional $16,240 is hardly a major problem. Conrads
real concern is the cost of the production tooling, the sales and marketing
collaterals, and the advertising campaign before, during, and after PRD. The
modied budget is approved.
John is relieved that things went well, but cannot help think that all
these extra meetings with top management require many hours of three or
four very expensive people. He is also acutely aware of the irony that the nal
development costs for Project Widget will probably wind up very close to his
original estimate. Had they not shaved 20% off the top initially, he would
actually be well within budget at this point, all these meetings would not have
been necessary, and he could have saved time by not having to divert his
attention. Oh well, apparently some things will never change.
The next day, Donna receives the RP&M model of the second design
iteration. It looks great, and the incremental weight increase is so small that
it is not really noticeable. She calls John and asks him if she should check
40
Jacobs
with someone in the model shop about having it painted to look just like the
nal design of the production Widget. John agrees and notes that if the painted
RP&M model can be ready by Monday, she should bring it to the project
meeting.
At the next meeting, Donna shows everyone the fully sanded, primed,
and painted model. The team is thrilled to see something that looks like a
product they could imagine people buying. Richard reports that with the approval of the budget increment, he has forwarded a purchase order to Central
Tool & Die for the rework. The schedule calls for the reworked tooling to be
ready in 6 weeks. Richard hopes Phil will complete the rst 20 functional
prototypes in about 5 weeks.
Jennifer shows the group the preliminary versions of the advertising
storyboards. The team likes them, although minor format changes are suggested by Ed. The health and safety package is about 85% complete, and the
certication forms for UL approval are ready to be submitted. UL testing will
be scheduled once functional prototypes are available. Sales brochures will
be prepared using photographs of the prototypes, after nal assembly.
The next 5 weeks seem to take forever. John double checks with Laura
and Andrew to be absolutely sure that all the thermocouples and strain gauges
are fully calibrated and ready for functional testing as soon as the parts arrive.
John also stares intently at a nondescript point on the wall while trying to
think of anything else that he could possibly do now that might save time
later. Precisely because the potential Widget market is signicant, he expects
that other people must realize this and may already be developing their own
version.
Just over 8 months ACG, the rst 20 functional prototypes in 35% glasslled polycarbonate arrive on Wednesday afternoon from Central Tool & Die.
John hands three of them to Laura and three more to Andrew. Within minutes,
the parts are being prepared for functional testing. By Friday afternoon, the
test results are rushed to an ad hoc Project Widget meeting called by John on
short notice. Finally, the test results all meet specication, Hallelujah!
The project team is elated. The prototypes look terric, the thermal results are better than before, and maximum deection at 120% of design load
is about 15% less than the specication. Operating the units as hard as they
can, even Richard, who was a linebacker on his college football team, cannot
induce enough deection to cause interference. Ed requests 150 prototypes
for his sales force, and Jennifer and George need about 200 for marketing test
samples.
Richard notes that the completely assembled tool, including the ejector
holes, ejector pins, ejector plate, cooling lines, registration pins, registration
41
holes, and of course, their precious Widget core and cavity set, is very heavy
and would take some time and money to deliver to another mold house. Furthermore, they would need to do all the paperwork necessary to cut a purchase
order for a new subcontract. If they stay with Central, the paperwork would
be easier and quicker. Also, because Centrals proof press could easily handle
another 350 parts, he suggests that they also complete the injection-molding
task. Besides, Central Tool & Die has really worked hard to deliver the prototypes as soon as possible, and they deserve the business.
This time, John and Richard visit the three local tool and die shops that
had previously bid on the prototype aluminum tooling. John shows them the
latest RP&M model, pointing out that the only signicant changes are the
10% increase in section thickness and the fact that 35% glass-lled polycarbonate has been selected as the production material. Because the nal geometry
is so close to the one they had bid on earlier, the three tool and die shops all
understand the project requirements. Richard asks all three to prepare formal
bid packages for production steel tooling sufcient to produce a minimum of
1.5 million parts in the rst year, 3 million parts the second year, and as many
as 4.5 million parts in year 3.
All three tool and die shops made it clear that this situation presents
some interesting alternatives. Nine million parts over 3 years pretty much
dictates the need for a multicavity tool or a very considerable budget for tooling rework. Because 35% glass-lled polycarbonate is highly abrasive, tool
wear will likely be substantial. If the market projection is correct, peak production will occur in year 3. The output in that year alone would strongly suggest
an eight-cavity tool. However, production the rst year would only require a
four-cavity tool. Richard realizes that numerous options exist, but which is
the best one? If only he knew what the Widget market was really going to be
like.
Phil at Central Tool & Die suggests to Richard that he could save some
money up front and hedge your market bet by going with a four-cavity
tool initially. This will almost certainly sufce for the rst year. If the Widget
market turns out to be less robust than expected, the four-cavity tool might
even see them halfway through the second year. Conversely, if the market is
booming, they will have bought some time with the rst tool and can always
purchase another four-cavity tool later. Obviously, Central will keep a copy
of the cutter path program, so the setup charges will be much less the second
time around. Also, if a year or so down the road they want to introduce a
Super-Widget, involving some product redesign, they can continue to produce
regular Widgets while the new tooling is being generated.
Almost as an afterthought, John also asks Phil for an estimate to injec-
42
Jacobs
43
44
Jacobs
Aiming toward the latter goal, lets review Project Widget with the intention of nding those tasks or procedures where time could have been saved.
In each case, we will (1) identify a specic event, (2) examine the result of
that event, and (3) propose a means by which the team could have saved time.
Clearly, all remedies will not apply in all cases, but if we can generate an
approach that provides substantially more than 28% savings in the product
development cycle, then all the proposed methods will not be required anyway.
1.
2.
3.
4.
45
Note that the rst four items are essentially cultural. They involve
using different operational strategies designed to save time. In this case, if all
46
Jacobs
four suggestions were utilized, the total time saved would be about 6 weeks,
or almost 11%. This is a good start. However, do not be deceived into thinking
that this will be easy. Changes in the way organizations do things is never
trivial. There is always the tendency to fall back on what has worked in the
past. However, it is no longer the past, and the competition is starting to play
smarter, and with better equipment.
5.
The team opts for CNC-machined aluminum prototype tooling, unaware of advances in rapid bridge tooling. From a CAD model, an
RP&M master pattern can provide composite aluminum-lled epoxy (CAFE) tooling. This method has been rened and improved
by RP&M service bureaus over the past 4 years and is used when
20500 prototypes are required in engineering thermoplastics. Another process is direct ACES injection molding (Direct AIM). The
core and cavity are built on a stereolithography apparatus (SLA),*
using the ACES (accurate clear epoxy solid) build style. Hand nishing of the master patterns is required for both CAFE and Direct
AIM. Unfortunately, female cavity nishing can take 2.53 times
longer than building the master pattern on the RP&M system! The
resulting core and cavity are mounted in a standard tool base [master
unit die (MUD), DME, National, etc.] and subsequently operated
on a plastic-injection-molding machine.
* Note: The machine is called an SLA, but both the process and resulting parts are
properly abbreviated as SL.
47
Note that if all six suggestions had been followed, the team could have
saved 17.5 weeks, or about 30% of their actual product-development cycle.
They would have reached PRD 2 months ahead of ACME, Inc., which would
have had a dramatic effect on their share of the new Widget market. Further,
they would have achieved all these time savings before they had even gotten
48
Jacobs
Had the team implemented all seven suggestions, the time savings could
have been a phenomenal 27.5 weeks, or almost half of their entire Widget
development cycle. This is not Fantasy Land. Multinational corporations,
original equipment manufacturers suppliers, and RP&M service bureaus using
rapid tooling are discovering time savings even greater than 50%. These dramatic reductions account for the growing interest in rapid tooling. The potential benets are enormous. Some forward-looking organizations have joined
consortiums to help them gain condence during early process renement (9
11). These companies know that once the techniques get past their growing
pains and mature into standard commercial practice, anyone NOT utilizing
rapid tooling will be at a serious disadvantage. Remember, as we said earlier,
to win in todays hypercompetitive global environment, you need to do some
things differently than the rest of the pack.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
49
plastic injection tooling. North American Stereolithography Users Group Meeting, Orlando, FL, 1997.
J Heath. Direct Tooling for Plastic Injection Molding. Proceedings of the SME
Rapid Prototyping and Manufacturing 96 Conference, Dearborn, MI, 1996.
P Jacobs. Recent advances in rapid tooling from stereolithography. 3D Systems
Report Number 70270, 1996, pp. 47.
S Rahmati, P Dickens. SL Injection Mold Tooling. Prototyping Technology International, International Press, Surrey, U.K., 1997, pp. 172176.
G Tromans. Casting and tooling applications of stereolithography at Rover
Group. Proceedings of the North American Stereolithography Users Group Meeting, Orlando, FL, 1997.
Laboratory to Advance Industrial Prototyping (LAIP), Clemson University,
Clemson, SC.
Rapid Prototyping & Manufacturing Institute (RPMI), Georgia Institute of Technology, Atlanta, GA.
LASER-engineered net shaping (LENS), Sandia National Laboratory, Albuquerque, NM.
4
Rapid Soft Tooling and Rapid Bridge
Tooling
Paul F. Jacobs
Laser FareAdvanced Technology Group
Warwick, Rhode Island
I. INTRODUCTION
Strictly speaking, the designation rapid modeling should have been utilized
to describe the various layer-additive technologies instead of the more commonly used term rapid prototyping. In agreement with standard manufacturing terminology, a model is an item which conveys the general shape of
something (i.e., the form of the object), as well as the nature of how it integrates with others as part of an assembly (i.e., the t of the object). However,
a model does not typically provide trustworthy information regarding the function of the nal part, because a model is usually not made from the nal
production material and is almost never generated using the nal production
method.
By denition, a true prototype is an object produced in the intended
material, by the nal method of production. For components ultimately to be
produced in metal, this might involve sand casting, investment casting, or die
casting. For products to be manufactured of plastic, the most common processes are injection molding and blow molding. By this denition, none of
the present rapid prototyping and manufacturing (RP&M) systems produce
true prototypes directly. Their are two obvious reasons for this. First, none of
the existing commercial RP&M systems can directly generate components in
aluminum or 35% glass-ber-lled polycarbonate, as our friends needed for
52
Jacobs
their Widget (see Chapter 3). Second, even allowing for technological advances, it is very unlikely that any of the existing RP&M systems, or even
those currently under development, will include investment-casting slurries,
die-casting equipment, or plastic-injection-molding capability as part of their
regular operation. Consequently, even if the various RP&M systems were able
to produce an extensive array of end-use materials, true prototypes would still
not be formed because, by denition, the nal production method would not
have been used.
The ability to perform part visualization, geometric verication, rapid
iteration, and form optimization was certainly key to the early growth of the
rapid prototyping industry (1). However, as noted earlier, this was actually
rapid modeling. When the various RP&M systems can build masters possessing the accuracy and surface nish required for tooling, only then will they be
capable of delivering true rapid prototypes in conjunction with an appropriate
secondary process.
For stereolithography (SL), this has already occurred to some extent.
Specically, over 25,000 QuickCast patterns have already been converted
into functional metal prototypes by means of investment casting (2). Also, SL
accurate clear epoxy solid (ACES) masters are currently being utilized in the
3D Keltool process to develop core and cavity inserts for plastic injection
molding (3). With some of the recent developments in rapid tooling, the M
in RP&M, has nally started to become noteworthy.
II.
RTV MOLDING
The most widely used form of rapid tooling currently involves silicone RTV
(room-temperature vulcanizing) molds. Of the roughly 300 RP&M service
bureaus currently operating worldwide, about half now provide RTV softtooling capability (4). Some large corporations have also installed RTV equipment internally to produce exemplars of some of their latest proprietary products. In the case of the defense industry, various classied military equipment
has also been evaluated in this manner. In these situations, the services of an
external bureau not holding the appropriate security clearance are effectively
unavailable.
The good news about RTV soft tooling is that it is very fast (e.g., some
service bureaus can provide a rst polyurethane part from an existing computeraided design (CAD) le within 5 days). RTV soft tooling is also substantially
less expensive than computer numerically controled (CNC)-machined aluminum tools. The bad news is that RTV soft tooling cannot generate true proto-
53
types, as the process yields only vacuum-cast polyurethane objects. Furthermore, these parts cannot be produced by injection molding, because RTV
molds are quite exible and would deform signicantly under the requisite
injection pressures.
The RTV process begins with a master pattern in the positive form
of the nal part. The master can be generated by hand carving, manual machining, CNC machining, and so forth. However, to save time, master patterns are
often built using RP&M techniques. Indeed, masters have been successfully
generated for RTV soft tooling from (a) fused deposition modeling (FDM),
(b) laminated object manufacturing (LOM), (c) selective laser sintering (SLS),
(d) solid ground curing (SGC), and (e) stereolithography (SL).
Other commercial RP&M systems, such as those from Sanders, EOS
GmbH, and Z-Corp. are also capable, at least in principle, of generating master
patterns for RTV soft tooling. Whether patterns generated by these various
technologies are indeed capable of the requisite master-pattern accuracy and
repeatability has generally not been convincingly established. Statistical process control (SPC) data involving a range of pattern geometries will be required from each of these technologies before this author, and presumably a
large number of potential users, will be satised.
The primary requirements for RTV soft tooling are that the master
Shall not cause RTV cure inhibition
Must possess the dimensional accuracy appropriate for the application
Should be able to be sanded/polished to the required surface nish
The latter point is important, because RTV is intrinsically capable of
replicating extremely ne detailsdown to the level of ngerprints left on a
glass microscope slide! Tiny aws on the master are picked up by the RTV
mold and subsequently transferred to the part. Interestingly, this is both a
blessing and a curse. The blessing is that RTV can, indeed, faithfully reproduce
ne detail, but the curse is that great care must be exercised during surface
nishing to avoid even the tiniest scratch appearing on the nal molded parts.
Prior to pouring RTV, a sprue is mounted on the master, typically with
a superglue. The sprue and master are wiped clean with a soft cloth moistened
with isopropyl alcohol, to remove dust and ngerprints. The master and sprue
are suspended in a clean wood or metal-forming box. The liquid-silicone RTV
material is mixed under vacuum to eliminate air bubbles. It is then poured
into the box and over the master while still under vacuum. The assembly is
then placed in a low-temperature curing oven and maintained at about 50C
(122F) for roughly 612 h.
54
Jacobs
Figure 1 RTV soft tooling of a hand-held scanner by Accent On Design, for Compsee, Inc.
55
registration. Compsee estimates that the use of RTV soft tooling on this project
saved approximately 75% of the time and 50% of the cost of this development
relative to the conventional methods previously employed.
The Compsee scanner was molded from one of the various two-part
polyurethane resins successfully vacuum cast in RTV molds. These polyurethanes provide a wide range of important mechanical properties, including (a)
hardness, (b) tensile strength, (c) tensile modulus, (d) exural strength, (e)
exural modulus, and (f) notched Izod impact resistance.
Table 1 lists these properties for three specic polyurethane resins, SG
95, SG 200 and 2170, distributed by MCP Systems. It also lists the same six
properties for ABS, Nylon 6, and polypropylene (PP). These data indicate that
recent advances in two-part thermoset polyurethane chemistry have provided
some interesting alternatives to the standard engineering thermoplastics.
On a decidedly larger scale, Fig. 2 illustrates another example of RTV
soft tooling. In this photograph, the ACES master is shown at the lower left,
the scalpel-cut RTV mold sections are at the top, and three vacuum-cast SG
95 polyurethane radio/cassette/compact disk boombox housings are located on the right. Each of the three housings was made from SG 95, but
different dyes were blended into the two-part polyurethane mix prior to vacuum molding. This enabled the generation of boombox housings in red, yellow, and black. A ballpoint pen is included near the model so that the reader
can get a sense of the size of these parts.
It is important to underscore two additional aspects of RTV soft tooling.
The rst is that solidied silicone has very poor thermal conductivity. Therefore, heat transfer from these molds can be exceptionally slow. For RTV mold
sizes up to roughly an 8-in. cube, 46 h are typically required before demolding the polyurethane part. If shorter intervals are attempted, inadequate dissi-
SG 95
SG 200
2170
ABS
Nylon 6
PP
Hardness (Shore D)
Flexural strength (kpsi)
Flexural modulus (kpsi)
Tensile strength (kpsi)
Tensile modulus (kpsi)
Notched Izod (ft-lbs/in.)
79
7.2
396
8.7
288
0.35
80
6.9
391
7.3
238
1.09
82
9.0
495
10.5
314
0.39
78
6.3
361
4.8
225
1.88
78
4.7
284
7.6
225
1.17
72
2.9
183
3.7
143
0.55
56
Jacobs
pation of the exothermic heat released during the two-part polyurethane reaction can, and often will, lead to poor part quality. For still larger RTV molds,
up to 12 h may be required.
Thus, although the generation of the RTV soft tooling may indeed be
exceptionally rapid relative to prototype aluminum tooling, the cycle times are
certainly not at all fast. If only three or four parts are needed, the added time
to mold and cure these parts will probably be less than 2 days. However,
as many as 23 additional weeks may be consumed in simply molding and
demolding 3040 parts.
Also, RTV soft tooling is indeed soft. The good news is that the exibility of cured silicone can greatly assist the demolding process. This is especially
true of those part geometries that involve minor undercuts. In these situations,
slide actions would normally be required on a conventional steel or aluminum
tool. Fortunately, an RTV mold can often be sufciently exed to enable part
release without the need to generate any slide actions, thereby greatly simplifying the mold design.
Unfortunately, the bad news is that RTV mold durability is marginal at
best. For very simple geometries without sharp edges, thin walls, or highaspect-ratio bosses, as many as 3040 parts may be obtained from a single
57
RTV mold. For moderately complicated geometries with a few sharp edges,
1530 parts of good quality can be anticipated. For highly complicated parts
with numerous sharp edges, extended thin walls, and multiple high-aspectratio pins or bosses, only 1015 acceptable parts can typically be produced.
Beyond about 15 parts, some portion of the RTV mold is likely to be either
torn or locally damaged.
Consequently, RTV soft tooling is best used when only a dozen or so
parts are needed, primarily as aesthetic models for photographs or as marketing test samples. Whenever the required part quantities increase beyond about
a dozen or the demands of functional testing with true prototypes become
critical, RTV soft tooling is probably no longer the proper choice.
Concept germination
Initial market assessment
Concept renement/denition
Competitive patent/legal status review
Generation of the initial product CAD design
Development of detailed functional product specications
Initial thermal, mechanical, electrical, chemical, or aerodynamic
analysis
8. Initial production cost/anticipated selling price estimates
9. Building a physical (possibly RP&M) model
10. Continued CAD design iteration
11. Form/aesthetic optimization
12. Initial FEA analysis
13. Identication of any potential design problems
14. Modications to the decient CAD design
15. Additional detailed FEA analysis/results
16. Development of prototype tooling
17. Generation of true prototype parts
18. Initial prototype functional tests
19. Additional CAD design changes
20. FEA analysis of the latest design
58
Jacobs
21.
22.
23.
24.
25.
59
60
Jacobs
As seen in Figs. 3 and 4, two key tasks, steps 16 and 21, each require
far more than 510 days to accomplish. This is the reason these two tasks
have been shown in boldface type. For Project Widget, which was intended
to be representative of a typical product-development scenario, step 16,
required 12 weeks to complete. Furthermore, because a problem with excessive deection occurred, step 21 was necessary, taking an additional 5.5
weeks. The total time required for these two tasks alone was 17.5 weeks.
Adding the 26 weeks for serial performance of the other 23 tasks, the total
shown in Fig. 3 is 43.5 weeks.
Note that for this case, the development and reworking of the prototype
tooling consumed 17.5/43.5, or about 40% of the time up to ordering production tooling. If concurrent engineering was used, and many of the other 23
tasks were done in parallel, the time savings would have been about 6.5 weeks,
the total time prior to placing the order for production tooling would have
been 37 weeks, and the prototype tooling would have consumed 17.5/37, or
almost half the time to that point, as shown in Fig. 4.
Oh, by the way, careful review of the progress our friends made on
Project Widget will show that they had completed the 25 tasks up to but not
including placing the order for production tooling, in 8 months, 3 weeks, and
2 days ACG, or about 38 weeks! Clearly, they must have utilized concurrent
engineering to a considerable extent. Unfortunately, the Widget team did not
utilize rapid tooling.
Historically, the central problems regarding prototype tooling have been
time and money. How does one produce just 50200 parts in a production
material without spending a lot of money and taking a lot of time? This has
been a very real dilemma for tens of thousands of companies working on the
development of millions of products. Until recently, none of the approaches
were very efcient, and all were quite expensive.
Traditionally, the most common procedure involved generating aluminum prototype tooling. Although aluminum can be CNC-machined more easily, rapidly, and economically than production steel tooling, neither the time
nor the money saved are enormous. If only 20 or so functional prototypes are
needed for mechanical or thermal testing, it is difcult to amortize the prototype tooling over such a tiny number. Fifty thousand dollars for 20 plastic
parts?
Nonetheless, many companies take the CNC/aluminum prototype tooling route. The good news is that true functional prototypes can be tested to
reveal potential problems with the product. The bad news is that this step is
both expensive and time-consuming, so time-to-market is further extended.
61
The second approach is to dispense with prototype tooling and prototypes altogether. The good news here is that considerable time and money are
saved, and the PRD can be moved forward signicantly. The bad news is that
the product may contain aws resulting in myriad failures, furious customers,
damage to the corporate reputation, and, in some cases, even protracted and
potentially onerous product liability lawsuits.
Some companies have tried a compromise approach, utilizing RTV soft
tooling to quickly generate polyurethane prototypes that are reasonably
close to the nal product. The good news is this method, as discussed earlier,
will denitely save time and money relative to the CNC/aluminum prototype
tooling approach. The bad news is that although better than nothing, these are
not true prototypes, and any test results based on their mechanical or thermal
properties will not be fully trustworthy relative to the nal product.
Fortunately, there is now a fourth option: Rapid Bridge Tooling. The
term bridge tooling was chosen to suggest that this approach can bridge the
gap between RTV soft tooling and true production tooling. The object, simply
stated, is to provide 20500 injection-molded prototypes in the desired production material, quickly and inexpensively. This is exactly what John, Laura,
Andrew, and Richard needed while working on Project Widget. Had they employed rapid bridge tooling to obtain just ve injection-molded prototypes in
glass-lled polycarbonate, the group could have accomplished the following:
1.
2.
3.
4.
5.
6.
IV.
Currently, there are three primary approaches to bridge tooling. The rst, and
most widely used, is composite aluminum-lled epoxy (CAFE) tooling. Many
service bureaus have been generating CAFE bridge tools for the past few
years. As an example, Laserform, Inc., located in Auburn Hills, MI (previously
62
Jacobs
part of Plynetics Express) had already built more than 150 successful, watercooled CAFE bridge tools (7).
From their experience, CAFE tools typically require between 3 and 5
weeks, have been made in sizes from 1 to 36 in., and can produce from 50
to 1000 parts, at a cost from $5000 to $20,000. Obviously, the cost and time
depend on the mold size and complexity. Also, the number of parts that can
be injection molded is strongly inuenced by the specic thermoplastic to be
molded and whether it is glass lled or not. Laserform had done up to 1000
polystyrene functional prototypes from a single CAFE tool, but as few as 50
parts in 40% glass-lled nylon (7).
A CAFE tool is typically generated directly from a positive master. The
master can be made in a variety of ways, including CNC machining of aluminum, plastic, or wood. However, Laserform preferred to utilize SL masters to
save time. The accuracy of SL masters is also constantly improving.
Figure 5 shows the continued reduction in the root-mean-square (RMS)
error for the stereolithography process. Data for the accuracy diagnostic test
Figure 5 Stereolithography UserPart RMS error versus time, from 1989 to 1997.
63
part known as the UserPart are plotted as a function of time from 1989 to
1997.
The UserPart was developed in 1989 by Dr. Edward Gargiulo, in conjunction with the North American Stereolithography User Group (8). It was
designed to establish the accuracy and repeatability of the SL process. This
diagnostic part measures 9.5 in. 9.5 in. in the XY drawing plane, by 1.5
in. in the Z or vertical direction. For each UserPart, 78 measurements are
made in the X direction, 78 in the Y direction, and 14 in the Z direction. The
measurements range from 0.125 to 9.500 in. Each point in Fig. 5 actually
represents a minimum of 1700 separate physical measurements, from at least
10 different UserParts.
It is clear from Fig. 5 that SL accuracy has improved with advances in
process, resins, software, and hardware. UserPart RMS errors of a properly
calibrated SLA-250, using Ciba epoxy resin SL-5170, are now under 45 m!
Pattern accuracy is one of the key prerequisites for tooling. It is not a coincidence that rapid tooling is gaining momentum as RP&M masters achieve this
level of accuracy.
Building a CAFE tool typically starts by sanding and polishing the master pattern. This is done to eliminate stair-stepping due to the nite layer
thickness used in all RP&M processes, as well as any other surface imperfections. Remember, the surface nish of the injection-molded prototypes will
only be as good as the nish on the master. Achieving the desired surface
nish is not trivial and can easily account for 2030% of the entire time required to make a bridge tool. Currently, most of this time is spent eliminating
stair-stepping artifacts. Building tooling masters with thinner layers to reduce
stair-stepping requires a longer period on the RP&M system, due to the additional overhead time associated with each layer. Nonetheless, the time saved
in reduced sanding is likely to be much greater. Furthermore, part accuracy
will be improved, as overzealous sanding can extend below the desired CAD
surface and ne features may be damaged or destroyed.
The sanded and polished master pattern is then coated with a thin lm
of a commercially available mold-release agent. The master is next accurately
registered inside a chase box. At this point, a parting surface must be selected.
If this is a plane, a simple wooden parting board can be used. However, if
the parting surface is more complicated, then a machined parting board is
appropriate.
Conformal cooling can be included by bending thin-wall copper tubing
and locating it inside the chase box near the master. The tubing geometry can
either be determined heuristically, or by means of a thermal nite-element
64
Jacobs
analysis (T-FEA) performed directly on the original solid CAD model of the
core or cavity. Good cooling will benet as follows (a):
1.
2.
3.
4.
65
as many as 5000 plastic parts can be produced from a single CAFE tool. They
have already achieved as much as 85% lead time reduction when employing
CAFE bridge tools relative to conventional CNC/EDM-generated tools. In
some cases, product-development cycles have been cut by a full year!
Furthermore, Kodak is typically saving about 25% in tooling cost. They
are also able to rapidly (a) test, (b) iterate, (c) retest, and (d) proof multiple
designs in less time that it previously took to just test a single design. Most
importantly, Kodak can now properly evaluate form, t, and function with
true prototypes injection molded in the desired end-use thermoplastic.
To date, over 40 CAFE molds have already been constructed and operated at Kodak. According to John Fowler, Supervisor of the Plastic Development and Fabrication Model Shop, SL masters combined with composite
aluminum-lled epoxy tooling have cut the time required for simple low-volume production molds from 810 weeks to just 24 weeks; and for complex
molds from 2638 weeks down to just 6 weeks!
Figure 7 shows another case involving SL masters and CAFE tooling
from Europe. ERU Elektroinstallation GmbH, in Thuringia, Germany, manufactures electrical consumer products. ERU had less than 1 year to design
and test a set of universal, multicircuit, two-way-control, illuminated switches
involving 27 different plastic-injection-molded components. Test results were
needed prior to committing to production tooling.
Further, ERU needed the exibility of making design modications
based on marketing inputs regarding customer preferences. Finally, it was an-
66
Jacobs
Figure 7 CAFE bridge tools for ERU Elektroinstallation GmH, produced by Schilling & Partners, Engineering.
67
on the ISS must meet very stringent requirements. However, these items will
only be produced in very limited numbers. Thus, it becomes extremely difcult
to justify hard tooling, as its cost cannot be efciently amortized over such
limited production quantities.
Bob Davis, Steve Irwin, and a team of engineers and scientists at Hamilton Sunstrand had to deal with this dilemma in the development of two components intended to be used aboard the ISS. Only 26 castings of each part were
needed. The rst component was to be produced in Inconel. Here, all 26 parts
were directly investment cast using SL QuickCast patterns. For the second
component, 26 aluminum castings were needed. The rst six parts were investment cast in aluminum using QuickCast. The remaining 20 aluminum parts
were investment cast using wax patterns molded in CAFE soft tooling. The
CAFE core and cavity inserts were themselves generated from an SL ACES
pattern. The resulting savings in time and cost were substantial relative to the
use of hard tooling. Furthermore, all of the resulting 52 parts successfully
conformed to a demanding 100% dimensional inspection. As a result, Hamilton Sunstrand Space Systems International is now dedicated to using the SL
process and rapid soft tooling on relevant projects and is currently transferring
this knowledge to other Hamilton Sunstrand divisions.
It is worth reecting on the impact that CAFE tooling might have had
on Project Widget. Based on results achieved by Kodak and Shilling, it is
reasonable to conclude that a time savings of about 3 months could also have
been realized by the Widget development team. Furthermore, this schedule
compression occurred just through the prototype stage! Additional and very
substantial time savings are possible should rapid production tooling ultimately be utilized, as discussed in Chapter 5.
68
Jacobs
Figure 8 Direct AIM core and cavity inserts used by Xerox Corp., and two injectionmolded polystyrene switch actuators.
Figure 8 shows a Direct AIM core insert on the left and a corresponding
cavity insert on the right. These inserts were built by Xerox Corporation on
an SLA-250 with Cibatool SL 5170 epoxy resin using the ACES build style.
An internal Xerox customer required 100 polystyrene switch actuators in a
very short time. After evaluating various alternatives, Jeff Heath decided to
try Direct AIM. His team was able to injection mold the required 100 polystyrene parts just 5 days after the CAD design was completed!
Table 2 provides injection temperatures, pressures, and cycle times for
a number of important engineering thermoplastics that have been injection
molded in Direct AIM molds. The parameters have not been fully optimized,
but they have been used successfully by a number of practitioners.
Figure 9 is a so-called scatter diagram which plots measured data
for a key dimension (viz. a diameter) on 200 polystyrene parts injection
molded into a Direct AIM core and cavity. The inserts were held in a standard
master unit die (MUD) frame.
Table 2 Suggested Injection Molding Parameters for Use with Direct AIM Core
and Cavity Inserts
Parameter
LDPE
HDPE
PS
PP
ABS
1600
180
3.5
2300
220
4.5
2400
200
4.0
1900
205
4.0
3200
240
5.0
69
Figure 9 Scatter diagram (dimension versus shot number) for 200 polystyrene parts
injection molded in a Direct AIM tool.
70
Jacobs
times allow the hot plastic inserts to cool below their glass transition temperature, thereby greatly increasing their strength and modulus. An effective procedure involves opening the press after the injection-molded plastic part has
fully solidied, and then blowing cool air on the Direct AIM core prior to
part ejection. The forced convection air cooling of the core will noticeably
increase tool life. The use of any of a number of commercial release agents
on every shot is also recommended. Again, although this does indeed add a
few seconds to each cycle, the overall project time savings can be so considerable that the extra effort involving mold release is well worth the advantages
related to mold survival and reduced time-to-market.
The primary advantage of Direct AIM is that the core and cavity inserts
are generated directly on an SLA, with no secondary processes required other
than preparation for installation on an injection-molding press. However, the
shortcomings of Direct AIM are as follows:
1.
2.
3.
4.
To address these issues, a number of variations of the Direct AIM concept have been developed and tested by 3D Systems, as well as a growing
list of users. In essence, these ideas involve different types of backing and
fronting materials. The rst of these variations is illustrated in Fig. 10.
71
Figure 10 Illustration of the concept of shelling a Direct AIM insert for backing
with aluminum-lled epoxy.
As shown in Fig. 10, rather than building a solid ACES core and a
solid ACES cavity, two relatively thin ACES shells are built on the
SLA. The benets are as follows:
1.
2.
3.
4.
72
Jacobs
2.
3.
The thermal conductivity values for the pure metals are 100 to 1000
times greater than those of the various composite bridge tooling
materials.
CAFE with 60% aluminum shot has a thermal conductivity that is
about an order of magnitude better than a straight ACES sample.
The compositeAIM sample (i.e., a 2-mm-thick ACES Direct AIM
shell, backed with CAFE mixed with 40% Al shot) had a thermal
conductivity about three times that of an ACES sample made from
solid SL 5170 resin.
From these results, it is clear that backing with appropriate materials can
improve the thermal conductivity of Direct AIM tools. The improved thermal
conductivity correspondingly reduces cycle time to about 2 min, from the
roughly of 45-min cycle times for solid direct AIM inserts. An extension of
this concept by Morgan (11) was the use of low-melting-point alloys of bismuth, tin, antimony, and lead as backing materials for Direct AIM thin shells.
Using the specic alloy CerroBend with a melting point of 58C, (136F)
to back a 1.5-mm-thick ACES shell built from SL 5170 resin, an effective
thermal conductivity for the composite article was determined to be about 8
103 cal/s cm C. This is over twenty times better than the thermal conduc-
73
tivity of a solid ACES insert of the same geometry, and almost 7 times better
thermal conductivity than an optimized compositeAIM insert. Here, cycle
times have been reduced to about 1 min.
An additional advantage of the low-melting-point alloy approach is that
when a tool has completed its function, the backing material can be melted
in boiling water and reused many times. With improved thermal conductivity
for signicantly faster cycle times, and enhanced compression strength for
longer tool life, recycling the backing material becomes even more cost-effective.
In addition to backing Direct AIM shells, a number of organizations are
currently studying various methods of fronting the active tool surface with a
material, or materials capable of the following:
74
Jacobs
1.
2.
3.
VI.
75
Form, and, nally, (c) polymer binder precoated low-carbon-steel particles for
RapidTool (14).
The RapidTool process is directly relevant to bridge tooling. It uses a
50-W CO2 infrared laser emitting at 10.6 m. The focused laser spot is scanned
with a pair of orthogonal mirrors. The energy absorbed from the moving laser
spot selectively fuses a thermoplastic polymer binder which has been precoated at a thickness of approximately 5 m onto low-carbon-steel particles.
The laser-fused binder holds the quasispherical steel particles together as a
green part. In this state, the part is rather fragile, with a green strength of
only about 440 psi, so care must be exercised to avoid damage to thin sections
during handling. The quasispherical low-carbon-steel particles have a mean
size of 55 m, with a size distribution extending from about 30 m to
roughly 75 m.
The green part is then placed in an electrical resistance furnace. Using a
25% hydrogen/75% nitrogen reducing atmosphere, the binder is almost totally
eliminated when the furnace temperature reaches about 700C ( 1300F).
The primary reduction product is methane (CH 4). Nitrogen and excess hydrogen will also exit the furnace. For environmental reasons, it is best if the CH 4
and any excess H 2 are passed through an afterburner, to enable combustion
in the presence of abundant ambient air.
At elevated temperatures, the nal combustion products will be primarily carbon dioxide and water vapor. Because any combustion process is never
perfect, trace amounts of carbon monoxide and various oxides of nitrogen will
also be generated. As they are produced in very small absolute quantities and
can economically be mixed with large amounts of excess air, the nal concentrations of CO and NOx can be made sufciently small to satisfy even the
most stringent environmental regulations.
Furthermore, just as any oxidation process is never perfect, neither is the
reduction process used to eliminate the binder. Small amounts of carbonaceous
residue will always remain and can actually act as a glue to temporarily
help hold the steel particles together. The small passageways that result from
the near elimination of the binder produce a porous article having about 60
vol% metal, and 40 vol% void space.
During the single secondary furnace cycle of the RapidTool process, (a)
the polymer binder coating is eliminated, (b) the steel powder is sintered, and,
nally, (c) the porous steel skeleton is inltrated with copper. The inltration
is accomplished by placing solid copper slugs on top of the green part prior
to the furnace cycle. When the furnace is heated to a temperature just above
the melting point of pure copper (1083C, or 1981F), but well below the
melting point of the low-carbon-steel particles, the molten copper then
76
Jacobs
wicks into the part through capillary action. The result is an essentially
fully dense part consisting of (a) the sintered low-carbon-steel particles and
(b) their interstices which have subsequently been inltrated with copper. The
total linear shrinkage occurring in the furnace is approximately 2.53.5% (14).
The primary advantage of the RapidTool process is that it forms a metal
part as its direct output, albeit after a secondary binder elimination/steel
particle sintering/copper inltration step. The resulting low-carbon-steel/
copper part can be used as a core or cavity insert for rapid tooling. However,
the elimination of the binder, the sintering operation, and the subsequent inltration process involve signicant linear shrinkage, as noted earlier. If this
shrinkage was absolutely constant for all geometries, at say 2.5%, it would
then be a simple matter to account for the entire process shrinkage by increasing the scale of all CAD dimensions by a factor of 1.025. Unfortunately, there
are two fundamental problems with this approach.
First, the shrinkage process is almost never perfectly uniform. In investment casting, sand casting, die casting, plastic injection molding, as well as
selective laser sintering, solid ground curing, fused deposition modeling, stereolithography, and, in fact, any process where there is a change of phase and
an accompanying shrinkage, careful experimental measurements invariably
show that thick sections will shrink somewhat differently than thin sections.
Second, as we shall discuss in the following section, is the issue of randomnoise shrinkage.
VII.
SHRINKAGE VARIATION
Many commercial processes involve a change of phase. Specically, a material may be transformed from a liquid to a solid, or in some cases from a solid
to a liquid and then back to a solid again. Some important examples are as
follows:
1.
2.
3.
4.
5.
Also, within the eld of RP&M, similar phase change phenomena occur. Specically the following:
1.
2.
3.
4.
77
In each case, the phase change, from liquid to solid, involves a decrease in
specic volume and a resulting shrinkage. The total volumetric shrinkage varies from process to process and from material to material, for a given process.
However, all of these processes involve some level of volumetric shrinkage.
Note that the volumetric shrinkage, Sv, and the linear shrinkage, S, are related,
for the case of perfectly isotropic shrinkage, by the expression
S 1 (1 S v )1/3
(1)
1!
2!
3!
(2)
1
Sv
3
(3)
Consequently, from Eq. (1), a material exhibiting 3% volumetric shrinkage should be undersize in all directions by [1 (1 0.03)1/3] 1
0.989898299 0.0101017 1.01017 %, provided the shrinkage was
perfectly isotropic. Note that by neglecting higher-order terms, the simplied
approximation of Eq. (3) would lead to a shrinkage of 1.00000% in all directions. Although Eq. (3) is certainly a close approximation, it is worth noting
that the difference between the two results, (i.e., 0.01017%) is hardly as trivial
as one might rst suppose when attempting to generate highly accurate core
and cavity inserts. For example, if we were working to develop a 20-in.-long
insert, the error associated with the approximation of Eq. (3) would account
78
Jacobs
for an error in the tool of 0.002034 in. all by itself. This is comparable to the
entire error budget for production tooling!
Furthermore, it is important to remember that a fundamental assumption
leading to this result was that the shrinkage is perfectly isotropic (i.e., identical
in all directions). Unfortunately, this is rarely the case for real parts! At the
atomic or molecular level, shrinkage may indeed be almost perfectly isotropic
(15). However, because cooling will always occur preferentially at the surface,
interior part temperatures will inevitably lag exterior part temperatures during
the cooling process. Consequently, shrinkage will tend to occur initially at the
outer perimeter of a part, and somewhat later within the interior of the part.
The result, even for a simple thin-wall section, would be slightly different
conditions acting on the central region of the part. This effect alone could
account for tiny variations in overall part shrinkage. Additionally, constrained
shrinkage associated with real part geometries (e.g., a thin-wall section joining
a thick-wall section) will also result in numerous small shrinkage variations.
Notwithstanding these issues, the basic approach used in all of the processes noted earlier involves some form of shrinkage compensation. Traditionally, one experimentally measures the linear shrinkage for a given material in
a given process, and then applies a shrinkage compensation factor to all part
dimensions. The part is intentionally built oversize, so that when the inevitable
process shrinkage occurs, the resulting part dimensions will be correct,
if the calculations have been done properly. This sounds nice in principle.
Unfortunately, experience indicates that it is not a simple matter to achieve
precise dimensional control through shrinkage compensation.
VIII.
BACKGROUND
79
80
Jacobs
IX.
RANDOM-NOISE SHRINKAGE
From experience we know that the shrinkage process is almost never perfectly
uniform. In investment casting, sand casting, die casting, plastic injection
molding, as well as SL, SLS, SGC, and FDM, and, in fact,any process where
there is a change of phase and an accompanying shrinkage, careful experimental measurements invariably show that thick sections will shrink differently
than thin sections.
Assume that we build a test part N times using (a) the same hardware,
(b) the same procedure, and (c) the same parameters, while holding (d) the
environmental conditions as constant as possible. Measuring the dimension
of each section and comparing this measurement with the intended CAD value
for that dimension, we can dene the linear shrinkage, S j,i for the ith measurement of the j th section, by the relation
S j,i
L j,CAD L j,i
L j,CAD
(4)
81
where L j,i is the ith measurement of the length (or width or height) of the jth
section and, L j,CAD is the intended CAD length (or width or height) of the j th
section.
Note that shrinkage is dimensionless, as it involves a length divided by
a length. We now dene the mean shrinkage for the j th section, S j, in the
usual manner, by summing the N separate shrinkage values, and then dividing
by N. In mathematical notation,where a bar over a quantity indicates the mean
value of that quantity,
Sj
1
N
j,i
(5)
i1
Figure 12 CMM measurements on a single section thickness for 30 stereolithography shrinkage test parts.
82
Jacobs
Figure 13
2.
There is also a small but denite variance between the mean shrinkage, S j, for each of the six different section dimensions.
3. Differences between the individual shrinkage values, S j,i, and the
mean shrinkage value, S j, for a given section are not trivial. The
desired, albeit not easily achieved tolerance for production rapid
tooling, is 0.002 in. on a 20-in. dimension. This corresponds to
an error of 1 part in 10,000 or 0.01%. To assure 99.7% of
dimensions will remain within this tolerance, three times the shrinkage standard deviation must not exceed 0.01%. Consequently,
the standard deviation of the shrinkage, s, should be less than
0.0033%, or about 1 part in 30,000. Tiny variations unimportant in
generating a rapid prototype for concept validation become critical
when attempting rapid tooling!
4. Finally, the difference between the mean shrinkage for one section
thickness, S j, and that for another section is also nontrivial at the
same level as that discussed in observation 3.
Two fundamental conclusions follow from Figs. 12 and 13 and the four
observations listed. The rst, embodied in observations 1 and 3, is that nontrivial differences in shrinkage occurring between otherwise identical sections,
of otherwise identical parts, prepared by identical people, in an identical manner, on identical equipment, using identical materials, under nearly identical
environments is a classic example of random noise!
83
The italics on the word identical are intended to remind the reader that
no two parts and no two experiments, and indeed no two measurements within
a given experiment are ever truly identical. In RP&M, tiny changes in temperature, pressure, humidity, laser power, and laser spot size will all effect the
outcome. With metal powders, the binder composition, particle size distribution, particle shape, part handling, and the details of the mixing procedure
have a nite inuence.
In short, it appears there will always be a component of random-noise
shrinkage superimposed on the mean process shrinkage. For convenience of
expression, let us dene the mean value of the shrinkage for a given process
by S and the standard deviation of that shrinkage by s. Note that because
shrinkage is itself dimensionless, then s is also dimensionless. We will return
to the matter of random-noise shrinkage shortly.
The second important conclusion from observations 2 and 4 involves
differences between the mean shrinkage, S, for different sections. The data
imply that shrinkage is fundamentally nonuniform. Indeed, the shrinkage is
dependent on the thickness of any section. Although the dependence is weak,
it is nite. Again, although a single shrinkage compensation factor is good
enough for prototype visualization, verication, and perhaps even iteration, it
is not sufcient when generating SL patterns for production rapid tooling, or
when using sintered powder metallurgy techniques for production core and
cavity inserts. The use of separate shrinkage compensation factors for dimensions in the X, Y, and Z coordinate directions is an improvement, but even
this approach does not account for variations in section thickness along the
same coordinate.
To achieve the accuracy levels required for production rapid tooling,
without postmachining, a more comprehensive method is required. The most
successful approach to date has been employed by some investment casting
foundries (23). It involves developing a body of experimental data for the
measured shrinkage values of a great many different shapes and then applying
slightly different shrinkage compensation factors to the CAD design for each
section of a part. The major advantage of this approach is improved part accuracy. The disadvantages are as follows:
1.
2.
3.
84
Jacobs
(6)
This concise relation assures that provided three times the standard deviation
of the shrinkage times the length of the relevant dimension is less than or
equal to the absolute value of the rapid tooling tolerance, then more than 99.7%
of all such dimensions should lie within that tolerance.
If this were the case, the process would at least be capable of reliably
providing accurate core and cavity inserts. If the mechanical properties, abra-
85
| T |
3L max
(7)
If we assume that L max 20 in. would sufce for the great majority of
rapid tooling applications, then the allowable standard deviation of the process
shrinkage would be
s
0.002 in.
0.000033 0.0033%
3 20 in.
Think about the implications of this result. If the mean shrinkage for a
process is quoted as 0.8%, but the actual value is really 0.79%, this would
hardly seem like a problem, right? WRONG! The difference between 0.80%
and 0.79% is obviously 0.01%. In the spirit of clarication, some simple numerical examples are as follows:
1.
2.
3.
4.
86
Jacobs
shrinkage uncertainty required for production rapid tooling. Unfortunately, the shrinkage for a given process would have to be quoted,
(e.g., as 0.037 0.003%). To the best of the authors knowledge,
none of the existing powder-metallurgy, based rapid tooling processes even begin to specify shrinkage at this low level, or with this
kind of statistical precision.
3D Keltool, RapidTool, Phast (24), and ExpressTools earlier powder
metal technique, all utilize some form of powder-metallurgy process. Each
involves a phase change and consequent shrinkage. How then could techniques
such as these ever hope to produce core and cavity inserts of sufcient accuracy to enable production rapid tooling without subsequent machining? The
answer lies in reducing the value of s. Of course, this sounds logical, but
how does one actually do this?
Since 1989, this author, as well as many co-workers, have been acutely
aware of, and deeply involved with the effects of shrinkage on the accuracy of
SL parts. Further, with the advent of the QuickCast process, similar problems
involving the effects of shrinkage on the dimensional accuracy of investment
cast parts also became evident. Additional studies, as well as discussions with
experts from various other metal-forming and plastic-injection-molding disciplines indicated the existence of related problems, albeit at different levels of
shrinkage and distortion. Shrinkage-related errors also occur in sand casting,
die casting, injection molding, as well as all the RP&M techniques.
X.
Figure 14 Linear shrinkage versus part number for the large-shrinkage case.
87
88
Jacobs
Figure 15 plots similar data for a different set of sintering materials and
process parameters. This time the mean process shrinkage is S 0.402% and
the standard deviation of the shrinkage is s 0.038%. Note that not only is
the mean process shrinkage smaller for this case, but the standard deviation
is also smaller. Again, taking the ratio s /S, we obtain the value 0.095, which
is very close to the previous result.
Finally, Fig. 16 plots similar data for yet another set of sintering materials and process parameters. This time the mean process shrinkage is S
0.201% and the standard deviation of the shrinkage is s 0.019%. Again,
taking the ratio s /S, we obtain the value, 0.095, which is also very close to
the results for the two other cases.
These data will support a hypothesis regarding shrinkage variation. Additional data are required for conrmation. Based on information available to
the author and the results presented in Figs. 1416, the following hypothesis
is proposed:
For processes involving a phase change, the resulting random-noise
shrinkage is directly proportional to the mean process shrinkage.
Mathematically, this statement takes the simple form
s KS
(8)
2.
3.
As noted earlier, when L L max, the equal sign applies in relation (6).
Substituting for s, from equation (8), we obtain the important result
SL max
| T |
3K
(9)
89
Figure 15 Linear shrinkage versus part number for the intermediate-shrinkage case.
90
Figure 16
Jacobs
91
Figure 17 Mean linear shrinkage versus length for various values of rapid tooling
tolerance.
92
Jacobs
For a given process and tolerance, the right side of Eq. (9) is a constant.
Thus, the mean linear process shrinkage, S, and the maximum tooling insert
length, L max, capable of satisfying the required tolerance, T, are related hyperbolically! This is shown in Fig. 17 for the following four values of the rapid
tooling tolerance, T : 0.002 in. for production injection-mold tooling,
0.005 in. for production blow-mold tooling, 0.010 in. for bridge tooling,
and 0.020 in. for soft tooling.
It is quite clear from Figure 17 that until the mean process shrinkage is
reduced to values less than 0.1%, the resulting random-noise shrinkage will
make it very difcult to achieve 5-in. production rapid tooling dimensions on
a consistent basis. Higher mean process shrinkage levels typical of current
powder metallurgy processes allow some dimensions to satisfy the respective
tolerances, but, unfortunately, others will not. Consequently, reliable rapid
tooling inserts capable of achieving production injection-molding tolerances
can presently be generated only for relatively small inserts.
Three potential solutions involve the following:
1.
2.
3.
Three such approaches are described in this book. One Involves nickelvapor deposition and two involve electroforming (CEMCOM and ExpressTool). Because electroforming involves essentially zero mean process
shrinkage, the problems associated with random-noise shrinkage are avoided.
Furthermore, the ExpressTool process has demonstrated 2030% cycle time
reductions relative to CNC-generated tooling through the use of high-conductivity backing materials and conformal cooling.
REFERENCES
1.
2.
3.
C Hull, P Jacobs. Introduction to RP&M. In: P Jacobs, ed. Rapid Phototyping &
Manufacturing: Fundamentals of Stereolithography. Dearborn, MI: SME Press;
1992, pp. 411.
L Andre, L Daniels, S Kennerknecht, B Sarkis. QuickCast foundry experience.
In: P Jacobs, ed. Stereolithography and Other RP&M Technologies. Dearborn,
MI: SME Press; 1996, pp. 209237.
D Smock. New moldmaking systems slice art-to-part cycles. Plast World Mag
3842, July 1995.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
93
T Wohlers. Ten inventions that have forever changed product development. Seventh International Conference on Rapid Phototyping, San Francisco, 1997.
T Kerschensteiner. AMP Inc., A Simultaneous Engineering Case Study. In: P
Jacobs, ed. Rapid Prototyping & Manufacturing: Fundamentals of Stereolithography. Dearborn, MI: SME Press, 1992, pp. 371380.
P Smith, D Reinertsen. Achieving overlapping activities. In: Developing Products in Half the Time. New York: Van Nostrand, 1991, pp. 153167.
S Willis. Real materials: Fast rapid tooling for injection molds. Proceedings of
the Seventh International Conference on Rapid Prototyping, San Francisco, 1997,
pp. 112.
E Gargiulo. In-plane stereolithography part accuracy. Proceedings of the 1st European Conference on Rapid Prototyping & Manufacturing, Nottingham, UK,
1992.
S Rahmati, P Dickens. SL Injection Mold Tooling. Surrey, UK: Prototyping
Technology International, UK International Press, 1997.
B Bedal, H Nguyen. Advances in part accuracy. In: P Jacobs, ed. Stereolithography and Other RP&M Technologies. Dearborn, MI: SME Press, 1996, pp. 156
164.
W Morgan. Low melting point alloys as backing materials for Direct AIM
plastic injection tooling. North American Stereolithography Users Group Meeting, Orlando, FL, 1997.
M Wilson, M Yeung. J Rapid Prototyping Tech 2(1), 1996.
K McAlea, P Forderhase, U Hejmadi, C Nelson. Materials and applications for
the SLS selective laser sintering process. Proceedings of the Seventh International Conference on Rapid Prototyping, San Francisco, 1997, pp. 2333.
U Hejmadi, K McAlea. Selective laser sintering of metal molds: The Rapid Tool
process. Proceedings of the Solid Freeform Fabrication Symposium, Austin, TX,
1996, pp. 97104.
T Mueller, Plynetics Express, personal communication, 1998.
H Nguyen, J Richter, P Jacobs, In: P Jacobs, ed. Rapid Prototyping & Manufacturing: Fundamentals of Stereolithography. Dearborn, MI: SME Press, 1992, pp.
250254.
L Andre, L Daniels, S Kennerknecht, B Sarkis. In: P Jacobs, ed. Stereolithography and other RP&M Technologies. Dearborn, MI: SME Press, 1996.
R. Connelly. Rapid tooling for medical products using 3D Keltool. Proceedings
of the Rapid Prototyping and Manufacturing 97 Conference, Dearborn, MI,
1997, pp. 8999.
K Dillon, R Terchek. U.S. Patent 4,431,449, Feb. 14, 1984, (assigned to 3M
Corp).
D Glynn, P Jacobs. CMM Measurements of Patterns and Powder Metal Inserts.
ExpressTool internal report, January 1998.
K McAlea, P Forderhase, U Hejmadi, C Nelson. Materials and Applications for
the SLS Process. Proceedings of the Seventh International Conference on Rapid
Prototyping, San Francisco, CA, 1997.
94
Jacobs
22.
23.
24.
25.
26.
5
Rapid Production Tooling
Paul F. Jacobs
Laser FareAdvanced Technology Group
Warwick, Rhode Island
I. INTRODUCTION
We have discussed some of the ways that the Project Widget team could have
utilized either rapid soft tooling or rapid bridge tooling to signicantly reduce
time to market. Through concurrent engineering practices they were able to
save about 6.5 weeks; cutting a 63.5-week product development cycle down
to 57 weeks. Furthermore, had they employed some of the cultural changes
described in chapter 3, section I, they might have saved another 6 weeks,
thereby slicing their product development cycle down to about 51 weeks.
Nonetheless, even with these admirable efforts, Acme still would have beaten
them to the market by about 2 weeks.
However, had the Widget team simply used CAFE rapid bridge tooling,
they could have cut another 11.5 weeks off the product-development cycle,
shrinking the latter from 51 weeks down to just under 40 weeks. Note that
40 weeks represents about a 37% reduction in the product-development cycle
relative to 63.5 weeks without concurrent engineering, and almost a 30% reduction relative to a 57-week cycle with concurrent engineering! The use of
CAFE alone would have enabled them to beat Acme to the marketplace by
2 months! Further, they would have saved about $40,000. Clearly, the gains
96
associated with rapid bridge tooling are already quite signicant and account
for its accelerated utilization by industry during the past few years.
Notwithstanding all these benets, the really dramatic advance will occur with the widespread implementation of rapid production tooling, which
has been the long-term goal of both authors, as well as many others, since
the early days of stereolithography. Surely, whatever size the worldwide prototyping market may be (reliable estimates have proved curiously elusive), it
is clear that the equivalent market for tooling, as well as subsequent manufacturing of various components through injection molding, blow molding, die
casting, powder injection molding, and investment casting, is likely at least
two orders of magnitude larger.
Even if the techniques of rapid prototyping (RP) advance signicantly
in terms of accuracy, surface quality, materials, speed, and cost reduction, the
RP market could be expected to encounter the rst signs of saturation at revenues of about $700 million per year, with signicant slowing of market growth
at about $1 billion per year. Why then do so many people at numerous organizations continue to be bullish on the future growth of the rapid prototyping
and manufacturing (RP&M) industry? The answer lies in the M.
The world market for manufacturing the items noted above is so enormous that even if rapid tooling is only able to garner a small slice, the overall
market for RP&M could increase by more than an order of magnitude over
the next decade. Will this actually happen? The answer is probably yes. When
something is needed by many and when that thing can save considerable time
and money, the pressure to invent, develop, improve, and commercialize a
practical, working version becomes very great. At least 25 different groups are
currently investigating rapid production tooling. Will they all be successful?
Probably not. Will one or a few of them be successful? Probably! In the words
of the late sportswriter Grantland Rice, The race is not always to the swiftest,
nor the battle always to the strongest, but thats the way to bet!
II.
97
to allow for (a) any shrinkage involved in making the pattern, (b) the shrinkage
of the 3D Keltool process, and (c) the shrinkage of the nal injection-molded
plastic). Alternatively, the pattern may be in the form of a negative (i.e., shaped
like the core or cavity which will ultimately mold the nal part, but again
oversized to allow for the shrinkage values noted above). The pattern may be
produced in various ways, including machining, grinding, carving, or by one
of the various RP&M methods.
In fact, it was precisely the difculty in rapidly achieving an accurate
pattern that was a major stumbling block for 3M in the 1970s and the early
to mid-1980s. It is ironic that during this period, 3M had developed a technique
to rapidly generate production tooling inserts and electrical discharge machining (EDM) electrodes, but did not have access to a rapid source of accurate,
high-quality, repeatable patterns. Consequently, much of the time saved generating the inserts or electrodes was offset by the lead time required to obtain
accurate patterns. Coupled with some of the practical limitations of the Tartan
Tooling process itself and the fact that neither the tooling cycle time reductions
nor the cost savings were sufciently great to establish a robust rapid tooling
market at that time, 3M decided to sell the process in 1986. It is perhaps
doubly ironic that about the time 3M decided to sell, 3D Systems was being
98
incorporated to develop stereolithography, a process that would ultimately provide a rapid source of reasonably accurate and repeatable patterns!
As seen in Fig. 1, there are two different routes to a nal 3D Keltool
insert. One path is known as reverse generation because it returns to the customer a core or a cavity insert which is the reverse of the pattern or le provided. This procedure is shown on the right side of Fig. 1. The second approach, is known as same generation, because it returns a 3D Keltool core or
cavity insert from a pattern or le shaped the same as the core or cavity (except
for the process shrinkage noted previously). This route is shown on the left
side of Fig. 1. The reader will note that the reverse generation method requires
six distinct steps to achieve a tooling insert, whereas the same generation
technique involves only four separate steps.
Because additional steps always introduce entropy, any transfer process
is never perfect. Why then would anyone intentionally pick the process that
involves two extra steps? The answer is because it is much easier to sand and
polish a positive pattern than a negative pattern. Although this may seem like
a small point, it is not. Experience has shown that sanding and polishing complex negative pattern geometries, to remove stair-stepping artifacts, can be
very time-consuming. It is not unusual to hear stories of a negative stereolithographic (SL) tooling pattern taking 1.5 days to build on an SLA, but then
needing 45 days just to sand and polish!
Remember, we are trying to reduce the overall time-to-market. Spending
5 days sanding and polishing a pattern, to save two additional 3D Keltool
process steps that require a total of only about 2 days is not the way to win
at this game. An interesting strategy used by some customers is to send a
same generation pattern of the core (which is intrinsically a positive and hence
relatively easy to polish), in order to save Keltool process time, and a reverse
generation pattern of the cavity, which is therefore also a positive, to save
nishing time.
The next step involves creating an room-temperature vulcanized (RTV)
silicone rubber Positive-in-a-Box intermediate mold. As seen in Fig. 1, this
may involve either one step or three steps. As discussed in Chapter 4, RTV
molds can faithfully reproduce ne detail. Also, their high exibility enables
the removal of fragile green compacts with reasonable yield. Unfortunately,
as also noted earlier, the CTE of typical RTV silicone rubber compounds is
300 106 mm/mm C. This is roughly 20 times the value of typical tool
steels! For a mold having a maximum linear dimension of 500 mm (20 in.),
a temperature difference in the RTV mold of only 1C (1.8F) will result in
its linear expansion by 0.15 mm, or 150 m, or about 0.006 in.!
99
The reader will quickly recognize that this source of error alone is three
times the desired rapid-tooling tolerance for plastic injection molding. Furthermore, this example involved a temperature difference of only 1C! Consider
the effects of the inherent silicone rubber exotherm, or what happens when
somebody opens a door to the process room on a hot day.
The good news about making intermediate molds from RTV is that the
surface replication is excellent, and the exibility of RTV assists in the demolding process. The bad news is that without precise and costly process
temperature control, the intermediate mold may be larger than, smaller than,
or fortuitously equal to the size of the pattern. Unfortunately, these variances
may not be small relative to production rapid-tooling tolerances.
After the RTV Positive-in-a-Box intermediate mold has completely
cured (hopefully at the correct temperature), a special bimodal mix consisting
of (a) A6 tool steel particles and (b) tungsten carbide (WC) particles is blended
with (c) a proprietary binder. The blending process is performed with a hightorque, water-cooled sigma mixer. Although thorough blending is certainly
important for achieving a uniform consistency, the blending cannot continue
for more that about 10 min or the binder will begin to cure. This would greatly
increase the viscosity of the mixture and impede complete RTV mold lling.
The 3D Keltool bimodal particle size distribution includes one group
of nely milled WC particles, ranging in diameter from about 1 m to about
4 m, with a mean effective diameter D WC 2.5 m. The WC particles
are generally polygonal or granular in shape. The second mode consists of
signicantly larger, quasispherical A 6 tool steel particles, ranging from about
20 m (viz. not passing through a #600 mesh sieve) to about 38 m (viz.,
passing through a #400 mesh sieve). Their mean diameter, D A6, is approximately 27 m.
This combination of particles provides the following benets relative to
simply using spherical particles of a single diameter:
1.
2.
3.
4.
5.
100
6.
7.
8.
9.
Figure 2 schematically illustrates the benets of a properly selected bimodal distribution on the nal packing fraction, relative to using spheres of
a single diameter.
It can be shown that the maximum possible packing fraction, [FP,I ]max,
for close-packed spheres all having the same diameter (i.e., a monomodal
distribution) is given by the expression
[FP,I ]max
(32)
0.74
(1)
However, for a bimodal particle size distribution, the ratio, R*, is dened by
the expression
R*
DL ,S
DS ,L
(2)
Figure 2 The inuence of (a) monomodal versus (b) bimodal particle distributions
on nal packing fraction. (From Ref. 1.)
101
where DL,S is the smallest diameter of the large particles and DS ,L is the largest
diameter of the small particles. Provided R* 7, then the smaller particles
will be able to ow into the interstices between the large particles. This
small-particle ow condition is key to obtaining higher values of the packing
fraction. The effect of particle size ratio on the binary packing fraction is
shown in Fig. 3.
It can also be shown that the maximum bimodal packing fraction,
[FP, ]max, is then given by
[FP, ]max [FP,I ]max (1 [FP ,I ]max)[FP,I ]max
0.74 (1 0.74)(0.74) 0.93
(3)
Here, the rst term is the maximum packing fraction for the large
spheres alone, and the second term is the maximum packing fraction for the
small spheres in the remaining void space, provided they are small enough to
ow into the interstices between the large spheres. To obtain high packing
fractions, there are clearly advantages to using a bimodal distribution relative
to a monomodal distribution. Extending this to trimodal or higher distributions
would seem logical. However, resolution issues as well as economics denitely establish practical limits.
Figure 3 The effect of particle size ratio on binary packing fraction. (From Ref. 1.)
102
103
104
is well below 100%. Techniques involving the use of air pressure on the RTV
side and a vacuum on the green part side can help, but they are not universal
solutions for all the various geometries encountered. Each new geometry can
be an adventure unto itself.
After a properly demolded green part is placed in a reduction furnace,
the ambient air is evacuated, and the furnace is then purged with nitrogen.
For safety, the vacuum pumping/nitrogen purge sequence is performed twice
to assure that the oxygen concentration is below 0.01% (i.e., 100 ppm). Next,
a continuous ow of hydrogen is introduced into the furnace. The electrical
heater elements are subsequently energized, and the electronically controlled
furnace temperature, TF, slowly ramps upward. When TF reaches about 350C
(660F), the hydrocarbon binder begins to decompose.
The dominant reaction is the reduction of carbon in a hydrogen atmosphere. The result is the production of methane (CH 4) according to the chemical reaction
C 2H 2 CH 4
(4)
105
The cooled, copper-alloy-inltrated, WC/A 6 tooling insert is then removed from the furnace. Next, the base is milled at, eliminating excess
copper-alloy inltrant. The insert is then forwarded to inspection, where surface quality is assessed and critical dimensions are measured to assure that
the part meets specication. If one or more dimensions do not meet specication, they may be machined if practical, or in some cases, it may be necessary
to repeat the last two steps shown in Fig. 1. When all is well, the part proceeds
to shipping. To save valuable time, it is returned to the customer by any of
a number of next day airborne/courier services.
The major advantages of the 3D Keltool process are now listed. These
advantages are relative to other rapid tooling processes that are currently commercially available. With the exception of items 3 and 6, they are not necessarily advantages relative to computer numerically controlled (CNC) and EDM
core and cavity machining, which clearly set the current standards for production tooling.
1.
2.
3.
4.
5.
6.
As a notable sidelight to illustrate the importance of rapid time-to-market, during the rst quarter of 1997, roughly 80% of 3D Keltool customers
chose the 2-week superrush schedule, in spite of the added cost. In many
circumstances, such as those faced by the Project Widget team, saving a few
106
weeks of product development time is clearly worth far more to many companies than the incremental fees for rapid delivery of tooling inserts.
The primary limitations of the 3D Keltool process are the following:
1.
2.
3.
4.
107
Figure 4 CAD solid model of a part, and the core and cavity mold geometries resulting from a Boolean reverse of that CAD model.
upper right, and a corresponding Boolean reverse. The latter has been divided
by a parting surface to form core (left) and cavity (right) mold sections. Early
work on investment cast tooling was done by Denton (4) and is described in
detail in Ref. 5.
The core and cavity patterns are produced using an RP&M system in a
format suitable for the shell investment-casting process. An excellent example
is the QuickCast build style developed by 3D Systems in 1992 and released
commercially in 1993 (6). Metal castings are produced directly from the
QuickCast patterns. After any necessary secondary machining, registration,
and assembly operations are performed, the mold is ready for use.
Cast tooling is nding opportunities in a number of female cavity mold
applications as inserts for the following:
1.
2.
3.
4.
5.
6.
Die casting
Rubber molds
Blow molding
Permanent molds
Plastic injection molding
Wax injection molds for investment casting
108
In the last application, the technologies have come full circle. Here,
CAD is used to (a) generate a solid model of the part, (b) establish the parting
surface, and (c) perform the Boolean reverse that establishes the core and
cavity geometries. After shrinkage compensation, the resulting core and cavity
solid CAD models are used to develop .STL les. An RP&M system then
builds appropriate patterns of the core and cavity that will then be shell investment cast. The core and cavity patterns are investment cast, nish machined,
aligned, and assembled in a tool base. The investment cast tool is then used
to mold production quantities of wax patterns for investment casting.
To be competitive with conventional CNC and EDM mold-making practices, (a) the CAD design of the component and the mold, (b) the RP&M
pattern fabrication, (c) the investment casting process, (d) the nal machining
steps, and (e) the tool assembly operations must all be
Fast
Accurate
Economical
The design of the mold for investment-cast rapid production tooling
employs practices similar to those encountered in conventional mold making,
including the following:
109
Applicable shrinkage factors are used to develop a shrinkage-compensated solid CAD model of the nal component. It is this shrinkage-compensated digital model that is ultimately used to establish the Boolean reverse,
the parting surface(s), the core and cavity geometries, the .STL les, and,
nally, the RP&M core and cavity patterns that will ultimately be investment
cast. All features required in the nal cast conguration must be incorporated
into the shrinkage-compensated solid model of the core or cavity. This includes llets, rounds, identication numbers, part markings, customer and
foundry trademarks, and so forth. Any casting enhancements such as (a) grinding stock, (b) vents, (c) drains, (d) wax gates and runners, and (e) liquid metal
gates and runners are CAD modeled at this time.
The heart of the investment-cast rapid production tooling (INCRPT)
process is the development of an appropriate shrinkage-compensated solid
CAD model of the nal component. If this is done correctly, generating solid
CAD models of the mold core and cavity are relatively straightforward. A
block shape is dened in CAD, having extents in X, Y, and Z. These must be
sufcient to ensure that the shrinkage-compensated solid CAD model of the
component can t inside the block, with enough room to spare in all directions.
This is important because the nal core and cavity must be strong enough to
provide the required life of the production tool. Once this has been done, the
shrinkage-compensated solid CAD model of the component is subtracted from
the block. The result of this reverse should be the desired geometry of the
nal mold.
Various CAD vendors have developed software packages for the specic
purpose of efciently generating mold geometries. Parametric Technologies
Corporation has a product for this purpose called ProMoldesign. The examples shown in Fig. 5, as well as others in this section, were generated using
this software module. The designer is assisted through real-time feedback relating to potential part/mold locking conditions. The opportunity is then available to change the conguration of the nal component in an effort to simplify
the design of the mold or to create an additional core or insert in order to
accommodate part extraction.
Once the mold components have been dened, a determination is made
regarding those surfaces that require excess machine stock. Candidate locations are those where as-cast surfaces will not satisfy the functional mold
requirements for surface nish, atness, or dimensional accuracy. This may be
due to the variability of the pattern-making process, random-noise shrinkage in
the investment-casting process, or a combination of both. Typical examples
that require nal machining would be parting surfaces and shutoffs.
110
Not all mold components will be cast. Where a feature can be formed
by a process which itself is faster, more accurate, or less expensive, that
method should be used. An example would be a round core pin machined on
a lathe from standard bar stock. Here, the pin can be machined much faster
and at signicantly lower cost than it could be investment cast. Nonetheless,
even though the component is made by another process, it is still included in
the CAD solid model. This allows the mold to be operated and cycled
in a virtual manner.
Through computer simulation, it is now possible to assemble the
mold components, inject wax into the mold, and view the lling action.
One can also disassemble the mold in the correct operational sequence,
extract the solidied wax pattern, and then evaluate its features for completeness. Again, all of these tasks can now be accomplished digitally.
It is well known that the best time to catch an error is at the earliest
possible point of discovery. The virtual world provides an excellent opportunity to efciently uncover such problems before expensive and time-consuming hardware changes must be made.
Once dened, the individual CAD models of the mold patterns must
have their cast shrinkage evaluations performed and the relevant shrinkage-
111
compensation factors applied. For the case study which follows, the time required to accomplish (a) the CAD solid modeling, (b) the complete shrinkagecompensation factoring, and (c) all the associated process engineering tasks
up to, but not including, the generation of the RP&M patterns was just 5
calendar days.
Throughout this discussion, all time intervals will be given in elapsed
or calendar time. This is quite different from what might be referred to as
stopwatch time, where only the time required to accomplish specic actions
or tasks is counted, with any queue time between tasks being conveniently
omitted. Ultimately, it is the total elapsed time that really matters when faced
with a deadline; hence, it is calendar time that will be reported. Of course,
the calendar time will always be longer than the stopwatch time often reported
by others. Therefore, the results may not appear as dramatic. However, the
data are indicative of what a user can realistically expect. Furthermore, just
as it is important to reduce the time for each individual step, analysis of total
elapsed time will probably point out other intervals where time can also be
saved.
The core and cavity patterns for this case study were created using the
QuickCast build style, which allows cured photopolymer to function as an
expendable pattern for the ceramic shell investment-casting process.
QuickCast establishes the pattern geometry with a thin skin (1 mm thick),
supported by an interconnected quasihollow hatch structure. This build style
allows the SL pattern to successfully emulate the petroleum-based wax patterns used in conventional investment casting (7). Figure 5 shows one of the
QuickCast patterns used in this study.
The QuickCast process initially requires checking that the pattern is well
drained and free of internal, uncured liquid resin. After postcure, the next step
involves lling the vent and drain holes that were intentionally generated to
evacuate uncured liquid resin from within the pattern. Filling the holes can
be done either with investment-casting wax or photopolymer resin thickened
to a pastelike consistency using ne powder ground from previously solidied
photopolymer. The pattern should then be tested to ensure that no openings
exist. This is best done with about 45 psi (0.3 bar) of positive pressure,
followed by drawing a mild vacuum (0.7 bar absolute pressure). In either
case, any leakage indicates the presence of one or more holes.
Once the pattern has been properly sealed, it is then connected to its
associated gating system. The pattern and gating are subsequently encapsulated in a multilayer ceramic slip and refractory grain coating. This coating
is allowed to air-dry, after which the entire assembly is placed in a furnace
preheated to about 1000C (1800F). The original QuickCast 1.0 software
112
113
The time sequence for the fabrication of INCRPT, including investment casting the production components, was as follows:
1. CAD solid modeling of component, mold, and engineering
tasks
2. QuickCast mold pattern generation
3. Investment casting of the mold components
4. Incorporating related features into nal mold, assembly,
and test
5. Molding the required wax patterns, and investment casting
5 days
3 days
5 days
2 days
20 days
5 days
114
115
IV.
The advent of chemically bonded sand has brought a new term and a new
capability to the world of the foundry. The term is precision sand casting. As
the name implies, the process yields castings with ner surface nish, more
intricate detail, and signicantly higher-dimensional accuracy than previously
possible with conventional green sand casting. Chemically bonded sand can
replicate a surface quickly, accurately, and economically. This enables RP&Mgenerated tooling solutions that can satisfy especially time-sensitive cast-metal
requirements.
The compatibility of chemically bonded sand with SL ACES patterns
is enabling the production of as many as 1000 castings from a given conguration. This unique tooling approach has already been successfully applied
for both short-run prototype and long-run production requirements.
In its simplest form, the chemically bonded sand approach involves a
sand mixer that coats very ne sand particles with a catalyst. This operation
is accomplished in isolation from a second mixing operation that coats similar
sand particles with a binding agent. Then, the catalyst-coated sand particles
are brought together with the binder-coated sand particles in a high-speed
mixing cone. Here, the two types of coated sand particles come in contact in
a continuous stream. The output of the mixing cone is directed over a pattern
set. The combined sand mixture is then tucked and hand compacted against
116
the pattern, which is held by a rail set in the XY plane. Figure 9 shows the
mixing cone, with the resulting stream of mixed binder- and catalyst-coated
ne sand particles being directed onto an ACES pattern.
The catalyst-to-binder ratio establishes the available working time
of the sand before it takes an initial set. When rst mixed, the sand is very
uid and is easily directed into the cope (top) and drag (bottom) pattern boxes.
After a given amount of time, based on the sand volume and the catalyst/
binder ratio, the mold will exhibit sufcient strength to allow inversion and
pattern withdrawal without damaging the cured sand mold, provided reasonable care is exercised.
Once fully prepped, the two mold sections are closed against one another
and clamped with sufcient force to withstand the hydrostatic pressure of the
117
molten metal during the pouring operation. The binder holding the sand particles together, at the interface between the cast metal and the precision sand
mold surface, is subsequently broken down by the high temperature of the
molten metal. This results in a loose sand envelope adjacent to the casting.
This thermal debinding facilitates the removal of the casting and its associated gating system from the mold with minimal stresses being imposed on
thin or delicate cast sections.
Fortunately, the precision sand-casting process can produce quality sand
castings quickly. For many simple congurations, it has already been demonstrated that it is possible to close the mold, pour molten metal, cool down,
and remove the solidied metal casting within 1 h. Obviously, larger and more
geometrically complex parts may take somewhat longer.
Although the thermal debinding mechanism does greatly assist casting
removal, it does, of course, destroy the mold. In that sense, precision sand
casting is similar to investment casting: both processes provide one casting
per mold. However, it may take 510 days to create an investment casting
shell from a single pattern. Furthermore, the pattern itself is eliminated in the
investment-casting process. With precision sand casting, the same pattern may
be used over and over again, and the mold can be produced in a matter of
hours.
Virtual pattern making is not a term of the future, it is a fact now. The
ability to utilize the skill sets of a journeyman pattern-maker to guide the
construction of precision sand tooling through the computer is becoming less
rare. Designing a pattern in a CAD environment employs procedures similar
to those used in conventional pattern making; specically, determining the
parting surfaces, establishing the core prints, dening core boxes, and so
forth.
The tooling for the precision sand case study shown in Fig. 10. involves
a single impression cope and drag plate with its associated core boxes. All
tooling components were modeled in solid CAD, including major portions of
the gating systems.
From the solid CAD model, the primary parting surface is dened and
the CAD model is split. Part features that will be formed by secondary cores
are identied. Appropriate core prints are also CAD modeled for the respective
cores. The core print, as well as with the core itself are extracted from the
model as a single entity. This is illustrated in Fig. 11.
At this point, a core box can be modeled around the core and core print.
This process is repeated until all cored areas are described. The sand mold,
with all its cores in place, can be simulated in the computer. Finally,molten
metal can be poured in the computer simulation, and the resulting casting can be extracted and inspected.
Copyright 2000 by Marcel Dekker, Inc. All Rights Reserved.
118
Figure 10 Rapid tooling for precision sand casting, including the cope and drag
plate and associated core boxes.
Figure 11
119
Once the design has been achieved, the individual components are generated on an SLA using the ACES build style. All tooling components are
currently built using a 0.004-in. (100 m) layer thickness for maximum
surface resolution and accuracy. Nonetheless, the components still require
some benching prior to mounting and assembly, in order to eliminate stairstepping on inclined surfaces. When thinner layers or advanced techniques
such as meniscus smoothing become available, the improved surface quality of inclined or compound curved surfaces will greatly reduce the amount
of benchwork. In turn, this will further accelerate the entire process.
Precision sand casting requires no external packing, pounding, or tamping. Consequently, the ne sand/binder/catalyst mixture can be molded
against an ACES part with very little abrasion. As a result, there is almost
no degradation of the ACES patterns during the sand-lling, mold-curing, or
pattern-extraction steps. A seal coat of paint applied in a light color is suggested to further aid in the visual inspection of abrasion on the active tooling
surfaces.
The ACES patterns have proven to be extraordinarily robust when used
in a production mode. Some congurations have yielded over 1000 precision
sand molds without any signs of wear. Obviously, care must be used in molding, pattern extraction, and general handling to allow for the reduced strength
and impact resistance of cured epoxy resins relative to either aluminum or
steel tooling. Experience to date indicates that tools fabricated in this manner
certainly require care in their use, but, of course, this is true for any precision
tooling.
For the case study described herein, the sequence of events and the time
required to develop Precision Sand-Cast Rapid Tooling is listed. Note that
this total elapsed calendar time includes not only the rst article production
casting, but weekend time as well.
1. Generating a solid CAD model of the casting from 2D
customer data
2. Solid CAD modeling of the tooling, as well as associated
engineering
3. Building the ACES patterns/core boxes
4. Bench nishing and assembling the tooling components
5. Producing the rst article casting and performing QA
inspection
Total calendar time from customer 2D data arrival until
delivery of the rst article casting
5 days
10 days
10 days
5 days
5 days
35 days
120
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
R German. Particle Packing Characteristics. Princeton, NJ: Metal Powder Industries Federation, 1989.
R McGeary. Mechanical packing of spherical particles. J Am Ceram Soc 44:
513522, 1961.
R German. Powder Metallurgy Science. 2nd ed. Princeton, NJ: Metal Powder
Industries Federation, 1994, pp. 242267.
K Denton, P Jacobs. QuickCast and Rapid Tooling: A case history at Ford Motor
Company. Proceedings of the SME Rapid Prototyping and Manufacturing 94
Conference, Dearborn, MI, 1994.
K Denton. Hard tooling applications of RP&M. In P Jacobs, ed. Stereolithography and Other RP&M Technologies. Dearborn, MI: SME Press/New York:
ASME Press, 1996, pp. 293315.
P Jacobs. The Development of QuickCast In: P Jacobs, ed. Stereolithography
and Other RP&M Technologies. Dearborn, MI: SME Press/New York: ASME
Press, 1996, pp. 183207.
L Andre, L Daniels, S Kennerkecht, B Sarkis. QuickCast foundry experience
In P Jacobs, ed. Stereolithography and Other RP&M Technologies. Dearborn,
MI: SPE Press/New York: ASME Press, 1996, pp. 209237.
P Blake, O Baumgardner. QuickCast applications. In: P Jacobs, ed. Stereolithography and Other RP&M Technologies. Dearborn, MI: SPE Press/New York:
ASME Press, 1996, pp. 239252.
R Hague, P Dickens. Stresses created in ceramic shells using QuickCast models.
Proceedings of the 5th European Conference on Rapid Prototyping and Manufacturing, Helsinki, 1996, pp. 1530.
6
Nickel Ceramic Composite Tooling
from RP&M Models
Sean Wise
CEMCOM Corporation
Baltimore, Maryland
I. INTRODUCTION
A matched die mold fabrication technique is discussed where nickel is electroformed over special tool mandrels made by rapid prototyping and manufacturing (RP&M) methods. The resultant nickel shells are then captured in a standard pocketed mold frame using a high-strength chemically bonded ceramic
(CBC) to secure the shell to the frame. The resulting nickel ceramic composite
(NCC) mold has a high-tensile-strength, abrasion-resistant surface, coupled to
the high-compressive-strength ceramic backing which provides support and
mechanical load transfer to the mold frame. The match of the ceramics thermal expansion coefcient to that of nickel, along with the net-shape forming
characteristics of both materials help maintain an effective bond and precise
location of the tooling components. This method was developed to produce
precise, high-quality fully functional tooling capable of intermediate volume
production runs in less than half the lead time of conventional machined metal
tooling. This chapter describes the tool-fabrication method through case studies undertaken as part of the development effort, as well as the molding performance of the tools.
122
II.
Wise
RAPID TOOLING
The model building industry has been revolutionized with the growth and
implementation of three-dimensional computer-aided design (3D CAD) tools
coupled with RP&M model building methods. In recent years, development
emphasis has shifted toward methods to create part-specic manufacturing
hardware such as tooling just as rapidly as plastic models. If this can be done,
then RP&M can become an integral part of the entire manufacturing process.
Additive processes are attempting to directly or indirectly produce such tooling. These include powder metal methods (1), cast metal (2), and metal deposition (3). In order to have a major impact on the very long lead items in an
original equipment manufacturers (OEM) product-development cycle, rapid
tooling methods must address the tooling needs for large parts, as this is where
the potential benets are greatest, and large tools are usually the pacing items
in product-development programs, as shown in Fig. 1.
If one considers that benching, tting, and nishing represent more than
a third of the fabrication time in conventional machined metal tooling, a truly
rapid process must produce an accurate tool that requires a minimal amount
Figure 1 Comparison of mold cost/lead time versus mold volume for tools made
via additive or subtractive methods.
123
of nishing or bench work. A truly rapid tooling process must therefore have
the following features:
1.
2.
3.
4.
5.
6.
One additive build method that can transfer geometry precisely from
plastic RP&M models is nickel electroforming. If this process is combined with
a high-strength backing material and standard mold components, tooling can be
produced that meets all or nearly all the criteria dened. The tooling system
presented in this chapter is called nickel ceramic composite tooling or simply
NCC tooling.
Figure 2 General layout of a NCC tool showing the use an electroformed face that
is coupled to a mold frame with rigid ceramic backing.
124
Wise
125
The difculty with a very stiff backing is that any dimensional changes
between the backing and the shell, such as those caused by differences in
CTE, or shrinkage of the backing after it has solidied, will result in a substantial shear stress at the interface. The COMTEK 66 chemically bonded ceramic,
with its CTE closely matched to Ni (13.9 ppm/C versus 13.5 ppm/C) also
provides very low shrinkage during cure (0.02%), resulting in an excellent
support material. (See Table 1.) Its room-temperature forming characteristics
are also important, as are its thermal conductivity. Although the ceramics
heat-transport properties are not as good as most metals, it is more than twice
as good as aluminum-lled epoxies [2.5 W/m K versus 1 W/m K (4)] and
ten times better than unlled epoxies [0.2 W/m K (5)]. In addition, the properties of the ceramic backing do not change when exposed to temperatures up
to 400C (750F) (6).
IV.
Recognizing the potential of the nickel ceramic combination, CEMCOM examined the use of RP&M mandrels as the geometric basis for this hard tooling
method. The process was used to form a test injection mold of an ice scraper
(Fig. 4) where the electroforming mandrel was an stereolithography (SL)
model. This work was performed in conjunction with Pennsylyania State University, Erie (3). The model was plated using a high-speed nickel process that
built up the required metal thickness in less than 4 days.
However, the electroforming conditions resulted in signicant deformation of the parting plane surface, and the nickel shutoff areas had to be ma-
Figure 4 Sketch of ice scraper part made in conjunction with Penn State University,
Erie.
Table 1
Tensile strength
Compressive strength
Flexural strength
Elastic modulus
Hardness
Coefcient of thermal expansion
Thermal conductivity
Specic heat
Shrinkage
Max. operating temperature
a
70 kpsi
29 mpsi
Rc 20
7.5 106 /F
468 BTU in./ft2 h F
0.11 BTU/lb F
Nil
500F
S.I. units
COMTEK 66
English units
S.I. units
50 kpsi
6 kpsi
5.4 mpsi
Rb 65
7.7 106 /F
17 BTU in./ft2 h F
0.19 BTU/lb F
0.2 mil/in.
400F ()a
350 MPa
42 MPa
37 GPa
13.9 106 /C
2.5 W/m K
787 J/kg C
0.02%
200C ()a
500 MPa
200 GPa
13.5 106 /C
67 W/m K
450 J/kg C
Nil
260C ()a
Above 200C (400F) shrinkage increases. COMTEK 66 has been used in plastic part fabrication at up to 315C (600F) and metal fabrication at
540C (1000F).
127
chined in order for the tool to close properly. Even so, once the parting surfaces
were tted, the tool was set up in a press and more than 1000 parts were
molded without signs of wear. This test showed the potential of the process,
but the stability and accuracy of the SL mandrel had to be improved if the
NCC process would ever satisfy the requirements for rapid tooling.
One way of avoiding the stability problems of the SL mandrel is to
only use it as the basis for a secondary pattern that is compatible with the
electroforming process. CEMCOM fabricated a tool in conjunction with the
Queensland Manufacturing Institute (QMI) and Marky Industries following
this process. This tool and the parts made in it are shown in Fig. 5. The electroforming mandrels were made from a standard tooling epoxy, then plated, and
the nickel shells subsequently separated. The shells were brought to CEMCOM,
backed with COMTEK 66 ceramic, and returned to QMI. The outer surfaces
of the ceramic were then machined at, tted to a base which contained alignment features, and the two halves were t together and nish machined to
make a close tolerance tool shutoff. The mold was then run to show that it
was capable of generating an injection-molded part. Although high-quality
molded pieces were again produced, the amount of tting and alignment was
very time-consuming. Excessive tting and alignment needs to be eliminated
if the process is to be fast and competitive.
Figure 5 Nickel ceramic composite mold tted to a mold plate along with molded
parts.
128
V.
Wise
Having established that the combination of nickel and chemically bonded ceramic works well, the focus of this project concentrated on integration of the
tool-forming steps with RP&M mandrel making (7). It is one thing to have a
process which is net shape relative to a model, but it is quite another to have
one where the models geometry will not be compromised by the processing
to which it is exposed. In essence, the process requires that a piece of plastic,
which is not as stable or as accurate as a machined piece of metal under the
best of circumstances, provide the precise geometry needed in a high-pressure
forming mold. Earlier experience with RP&M mandrels used to make the PSU
ice scraper showed these limitations. The electroforming process takes place
under water, well above room temperature. As the nickel deposits on one side
of the model, it seals the surface to moisture penetration. If just one side of
the RP&M mandrel is being used to form the tool surface, then bending in
the pattern will take place as one side swells while the other side does not.
Once this happens on core and cavity patterns, the parting plane is compromised, so the tool is both difcult to seal and the part thickness is not precisely
controlled.
If the part model is designed so that the core and cavity geometry are
each attached to a common parting plane, then the tendency to distort will be
minimized because both will be sealed by the nickel. If any swelling takes
place, it will be very nearly identical on both sides. In addition, even if minor
distortion does take place, the connection between core and cavity ensures
that one side will follow the other. Besides assuring better parting line accuracy, the two-sided-model concept provides substantial benets with respect
to core and cavity alignment during tool assembly.
The process for using a two-sided RP&M mandrel as the basis for an
NCC tool is illustrated in Fig. 6. It begins with the electronic 3D solid CAD
model of the part. A part designer in conjunction with a tool designer analyzes
the CAD model, and the parting surface is dened around the perimeter of
the part. This surface, and the part model itself, is expanded in a linear fashion
in the mold opening or z direction. The amount of z expansion should be
chosen to provide a model with good stability for the plating conditions. If
the part is edge gated, the runner and gate should be laid out on the parting
plane to ensure that there is sufcient nickel over these surfaces. The parting
surface should also have provisions for alignment to the mold frame, as this
will facilitate assembly and backing in the later stages of toolmaking.
129
The particular model used in this test had a buss bar/clamp ring attached
to its perimeter. This ring had three purposes: (a) create electrical contact
around the perimeter of the model, (b) clamp the model to prevent bowing or
distortion around the edges, and (c) provide a means of attachment to the mold
frame. Because the clamp ring attaches to a machined recess in the mold frame,
these two rings need to be precisely aligned relative to one another, or a shift
will occur in the core relative to the cavity during nal assembly.
The nickel shells are electroformed over the mandrel using plating conditions that minimize dimensional changes of the RP&M material. The amount
of nickel needed on the model will be dependent on factors related to the
stresses incurred during part forming, the number of parts needed, and so
forth. With this method, the nickel thickness required is somewhat less than
ordinarily used for electroformed shells because the ceramic backing will provide the needed support in the thinner areas and the shell does not need to be
removed from the model until after it has been backed. This approach allows
shells to be made in days rather than weeks.
The RP&M mandrel, with the electroformed nickel on both front and
back surfaces, is attached to the mold frame using the clamp-ring buss bars.
130
Wise
The ceramic is then vacuum cast through a small opening in the back of the
frame. The material must harden overnight before the mold is ipped and the
opposing side can be cast. Once the opposing side hardens, the two halves
are separated, removing the RP mandrel from the core and cavity. This step
is then followed by a cursory inspection. Next, the ceramic is hydrothermally
cured and measured and the t of the core and cavity checked. The nickel
surfaces can be cleaned or polished at this stage if needed. RP models normally
have roughness and/or stair-stepping on vertical walls. Much of this should
be sanded smooth prior to electroforming, but there may be inaccessible areas
on the mandrel which are difcult to nish. Given the ability of the nickel to
pick up very ne detail, these imperfections transfer to the nickel surface. The
work required to clean this up is usually minimal, as the most recessed areas
on the models now stand proud on the tool surfaces. Sampling of the mold
can be performed without any nishing of the nickel tool face, but the rough
surfaces with a low draft angle may not release well. In order to get the tool
ready for molding parts, holes for ejector pins need to be drilled through the
nickel, ceramic, and mold frame. CBC is sufciently machinable that holes
for ejector pins larger than 1/8 in. in diameter can be drilled accurately. Holes
smaller than 1/8 in. in diameter may require a bushing. The runner is also
extended through the buss bar ring and into the mold frame to connect it to
the sprue, enabling the tool to mold parts.
VI.
The model build approach outlined in Figure 6 was tested on a part geometry
and SL mandrel supplied by Doug Van Putte formerly of Kodak (8). This
part, shown in Fig. 7, is a 25 51 25-mm (1 2 1-in.) rectangularbox-shaped piece with a 1.27-mm (0.050-in.) wall and a semicircular cutout
of 12.5-mm (0.5-in.) radius on one side. There are two ribs on one end of the
part, leaving details less than 6 mm (0.25 in.) wide but more than 18 mm
(0.75 in.) deep to electroform. This is difcult geometry for the electroforming
process. However, closely spaced ribs are commonly found on plastic parts.
The two-sided SL electroforming mandrel had a parting line thickness of 12.5
mm (0.5 in.) The part was edge gated so the runner and gate were built on
the parting line.
Field control devices were mounted over the part prior to placing it in
the electroforming tank. These helped create a more balanced current density
around the part surface. After about 1 week of plating, the core side had nickel
thicknesses that ranged from 0.6 mm in the deepest recesses of the pattern to
131
more than 2.5 mm at the parting line. The cavity side had even greater nickel
thickness plated over the highest part of the model. This is fortuitous because
the cavity of the tool sees the highest tensile stresses when parts are molded.
A bonding layer was applied to the back of the nickel shell and to the
mold frame. Normally, conformal cooling lines would be placed between the
shell and the steel frame, but given the size of the tool, it was decided to
simply rely on conduction of heat through the ceramic to the mold frame [a
master unit die (MUD) insert]. The buss bar clamp frame was then bolted to
the machined MUD base and the nickel shells were backed in sequential fashion then cured hydrothermally following the procedures outlined in the previous section. Some machining was done on the buss bar clamp ring because
it did not sit ush on the MUD frame. The nished mold halves shown in
Fig. 8 were then returned to Kodak for measuring and part molding.
Preliminary measurements prior to shipment, subsequently conrmed
by Kodak, showed that the parting line had distorted a small amount. The core
side was concave by 0.1 mm (0.004 in.) and the cavity side was convex by
0.2 mm (0.008 in.) This distortion is believed to be due to the restraint of the
plastic model at its perimeter by the buss bar clamping frame, coupled with
the differential thermal expansion of the metal and plastic that occurs at the
nickel electroforming temperature. Overall, it was observed that there was a
slight expansion of the core and cavity geometry relative to the SL mandrel
by 0.030.1 mm (0.0010.004 in.).
After the ejector pin holes were drilled for the knockout and the runner
was cut from the edge of the shell to the sprue, the tool was mounted in a
press and parts were molded from polystyrene using an injection pressure of
700 bar (10,300 psi). Steady-state cavity and core surface temperatures of
132
Wise
Figure 8 Kodak tool core and cavity NCC inserts in MUD frame. Molded parts are
shown in foreground.
50C (122F) and 60C (140F), respectively, were measured using a molding
cycle of 40 s. (A steel tool for the same part ran on a 30-s cycle.) This was
considered quite reasonable given the fact that no special cooling provisions
were included in the mold. In order to get the tool to run on a fully automatic
cycle, injection pressure was reduced to 380 bar (5600 psi). A total of 5000
high-impact polystyrene parts were molded. The tool did not show any wear
or deterioration from running these parts.
After injection molding, the parts were sectioned and the wall thicknesses were measured. They ranged from 1.2 to 1.5 mm (0.048 to 0.060 in.).
Nominal wall thickness was 1.3 mm (0.052 in.). The largest differences were
seen between the side with the gate and the opposing side that had the 0.5 in.
radius cut out. The gated side was thickest. It is not known if this is due to
differential pressure from one side relative to the other, or a slight shifting of
the core on assembly. The earliest parts run at a pressure of more than 700
bar had a small amount of ash in one corner, but this was eliminated when
the molding parameters were optimized for automatic operation.
When Kodak completed their durability test, the mold was returned to
CEMCOM Corp. so that it could be used for longer-duration durability trials.
133
Cycle time
Polystyrene
Polyethylene
PVC
Polypropylene
PBT 30% glass
Nylon, 30% glass and mineral
ABS 30% glass
Polycarbonate
Total no. of parts molded
a
40
30
30
30
22
15
25
30
sa
s
s
s
s
s
s
s
No. of parts
5,000
15,000
5,000
5,000
5,000
5,000
5,000
1,000
46,000
Prior to running these tests however, cooling lines were added to the mold
frame and the surfaces were dressed so that better release would be obtained.
This allowed the tool to run fully automatic on a faster molding cycle. Table
2, shows the materials, the cycle times, and the volume of parts molded from
each resin. As can be seen, the NCC mold handled both corrosive [poly(vinyl
chloride) (PVC)] and abrasive, glass-reinforced materials well. Electroformed
nickel is very corrosion resistant, so the performance with PVC was as expected. However, the nickel surface is softer than most tool steels. Whereas
molding 15,000 parts with glass-reinforced resin did not change any dimensions on the inserts, even around the gate, there was a noticeable polishing
of the nickel surface in this area. Sticking of polycarbonate to the core surface
ultimately caused the tool to fail, but more than 46,000 shots were run. This
trial demonstrates that the NCC system has the capability to injection mold
reasonable production volumes.
134
Wise
who needed a large internal part for one of their mailing machines. Figure 9
shows three views of SL tool mandrel for a part that is 385 125 85 mm
deep. With the addition of the extra material for the stepped parting line and
the clamping frame around the perimeter, a mandrel 480 230 100 mm
(19 9 4 in.) was made. The reader can see in these three views the detail
for the ejector side geometry on the left, the injector side detail on the right,
and the alignment and registration of these detail in the center view. A few
of the features on this part were narrow slots which would be difcult to
electroform. These small features were machined from metal and tted to the
mandrel prior to electroforming. They were then simply captured in the nickel
shell as it was formed. In addition, due to the large size of the part and the
need for high precision in the boss locations, it was decided that these would
be drilled and placed in the tool after the shell had been formed and captured
in the mold frame.
Figure 10 shows the NCC tool after demolding but prior to nal machining for the knockout system, pins, and the sprue bushing, which was located
Figure 9 Three views of the SL mandrel used to make mailing machine part for
Pitney Bowes.
135
Figure 10 Nickel ceramic composite tool prior to nal machining of the knockout
system. The ruler in front of the tool is 600 mm (24 in.) long.
near the center of the part in this case. Note the cooling lines emerging from
the left and right sides of the tool. These were cast into the ceramic. The
fabrication time from receipt of model to the point shown in this photograph
was just under 5 weeks. The nal machining and nishing brought the total
to six weeks. Some tting and bluing was required, particularly on the inclined
136
Wise
shutoff areas where a good deal of stair-stepping had to be manually removed from the SL mandrel. This left the nickel just slightly proud in these
regions.
Parts were molded on the tool using a 30% glass-ber-reinforced Noryl
PPO structural foam thermoplastic. Molding cycle was nearly 3.5 min., which
is 15% longer than the part would have run in steel. A photograph of the
front and back of the molded part is shown in Fig. 11. The part on the right
is turned to show the injection side. The white mark where the sprue has been
removed is visible. Perhaps the most notable feature of this large injectionmolded part is the fact there is no ash. The part came out of the tool very
cleanly, demonstrating the viability of the two-sided mandrel approach for
injection molding larger-sized parts.
The part was measured as a quality control check prior to tting to a
functional mailing machine. The holes in the bosses, that must line up with
Figure 11 Two parts showing the ejector side and injector side detail of the internal
part for a Pitney Bowes mailing machine. The ruler in the foreground is 30 mm (12
in.) long.
137
other features, were within 0.1 mm of their intended location. This is no surprise because these pins were tted to the tool after the forming process. The
overall length of the part was nearly 0.6 mm longer than the design length.
Given the thermal expansion coefcient of the SL mandrels epoxy photopolymer, this dimensional change is consistent with the expected thermal expansion of the mandrel at the electroforming temperature. In the future, a simple
thermal expansion correction of 0.13% of the mandrel dimentions during the
RP&M step should bring the size of the part within a still tighter range. Even
so, the part as made was within specications. It t and functioned well with
the other components in the system.
138
Wise
Because this was an appearance part which had a texture on the surface,
there was some discussion about applying texture directly to the model at the
beginning versus texturing the nickel surface after the tool had been sampled.
(The latter is standard practice in steel tools.) As it would be extremely difcult
to repair any aws in any textured area while assuring that this repair would
remain invisible in the molded part, it was decided that the appearance side
of the tool would be made smooth, and textured later. There was some concern
that texturing the nickel may be difcult due to the metals corrosion resistance. However, Moldtech was able to apply the desired texture using a standard mask with a stronger than normal etching solution.
Because tool shutoffs are built into the SIMPLE tool model, there is
high accuracy required in the region just outside the perimeter of the part.
Tolerances here need to be held to within 0.002 in. This kind of accuracy in
the build direction is difcult to achieve with the existing RP&M processes.
When coupled with a build layer thickness of 6 mils, the shutoff areas needed
a signicant amount of hand work on both the mandrel and the nickel shell.
Note also that in the hinge area, the parting surface steps down to the level
of the detail resulting in a very steep shutoff region.
Once the SIMPLE tool mandrel had been carefully sanded down to at
least a 600-grit nish, the features to be captured in the shell were mounted
139
Figure 13 Appearance part tool in press. The textured side of the mold is on the
left. Reverse ejection with slide action features are shown on the right.
to it, then coated with a conductive layer, placed in the plating tank, and
attached to a power rectier. The mandrel was plated for 10 days and, in this
time, nickel was built up to an average thickness of 0.150 in. Inside corners
and recessed areas had signicantly less nickel than outside corners, but the
ceramic backing lls in the unevenness so that the shell will be uniformly
supported in operation. The shells produced remained on the model while a
proprietary bonding layer was applied. Cooling lines were also mounted three
tube diameters behind the shell and special attention was directed to areas of
the core which were expected to be the hottest. The SL mandrel was then
used to locate the core and cavity in the mold frame, and the CBC backing
was applied. After the ceramic set for at least 24 h at room temperature, the
mandrel was removed and the NCC tool was postcured.
The core side of the NCC mold was then machined to accept the hot
sprue bushing, and holes were drilled through the nickel and the ceramic for
each of the ejector pins. The mechanical pulls for the slide actions were also
mounted to the frame at this time, in a fashion similar to a machined metal
tool except that they were aligned to the preformed holes in the shell and
140
Wise
hinge core inserts. Preliminary nishing was done on the nickel at this time,
which included quick bluing of the parting plane. The tool was then sent to
an injection-molding house to be sampled. The tool is seen in the molding
press in Fig. 13.
Minor ashing was found at the corners, and the hinge detail was difcult to ll without burning. Also, the shallow draft hinge area tended to hang
up in the tool as parts were run. To overcome these problems, the tool was
more thoroughly blued to obtain proper closure around the entire perimeter.
Some buildup was necessary and this was accomplished via microwelding.
Once the closure was correct, the cavity side was textured and the core detail
was more carefully nished around the hinges. Small ejector pins were added
to the lowest part of the hinge-forming area in the tool (primarily for venting),
and extra ejector pins were added adjacent to each of the hinges. With these
modications, the tool produced the parts shown in Fig. 14.
IX.
COMPRESSION TOOLING
The tool model designs discussed are well suited to simple closures of matched
die molds. There is signicant application for this technology in the compres-
Figure 14 Molded parts from the NCC tool. The textured side is on the left and the
injection/ejection side is on the right.
Copyright 2000 by Marcel Dekker, Inc. All Rights Reserved.
141
sion-tooling market where full positive closures are the norm. Expanding a
tool mandrel in the mold opening direction for this type of closure would
require that the tool model be substantially thicker and therefore more costly.
Additionally, the closure on the core side of these tools are more likely to be
eroded or degraded and it would be preferable that these elements be made
from a harder material than Ni. To overcome these problems, a method to
build aligned shear closures by mounting the mandrel on a machined steel
insert was developed. The insert ultimately becomes the core side shear edge.
This is shown in Fig. 15. Another feature in this tool is a welded plate mold
box rather than a standard mold frame. This can be done because low-volume
compression tools can often be run with poppettes rather than a full ejector
plate with pins. Also, compression tools are often aligned at the center of the
tool using heel blocks rather than pins placed in the corners.
This low-cost tool fabrication approach has been demonstrated on two
tools, in conjunction with Zehrco Plastics. The second of these, a two-cavity
mold for a sensor case, was made in 3.5 weeks from reciept of CAD data and
ran at standard operating conditions for a polyester bulk mold compound of
Figure 15
molds.
142
Wise
325F and over 1000 psi. The mold produced more than 1500 parts. Molds
using this design should work just as well for sheet molding compound. This
basic approach has also been used for tools to mold rubber.
X.
CONCLUSIONS
XI.
FUTURE WORK
Having demonstrated that the NCC is a viable mold-making process, the practical limits need to be explored for incorporation of slides and lifters. Their
incorporation will be tested in different stages of the NCC tool fabrication
143
process to see how they affect the overall fabrication costs and timing. The
impact of the design features outlined in the chapter will be assessed on the
overall fabrication time of the NCC tooling. It is imperative that methods and
process sequences developed do not extened the lead times possible with the
NCC tooling method. To this end, more effort will be directed at standardization and modularization so that the process speed can be further improved.
Finally, the part size envelope will also be explored. Components that are
nearly 0.5 m 1 m are under consideration for fabrication from RP models.
The process is suitable for even larger parts, but it may not be feasible to use
current RP model-making methods.
ACKNOWLEDGMENTS
The author gratefully acknowledges the contributions of the following people
and organizations to this effort: Doug Van Putte and his associates at Kodak
for providing the SL mandrels, the MUD frame, and for molding parts in the
small demonstration tool; Vadan Nagarsheth and Glen Randmer of Pitney
Bowes Inc. for continued support of the NCC tooling effort and their contributions to both part design, tool design, and CAD le preparation for the mailing
machine part; Mike Naylon of QMI for work with the phone insert part; and
Rob Tanis, Lee Robinson, and Kevan Jones of CEMCOM Corp. for their NCC
tool-fabrication efforts.
REFERENCES
1. W Durden. A successful team approach to rapid tooling. SME Rapid Prototyping
and Manufacturing Conference Proceedings, 1996. T Gornet. Experiences with
DTM RapidTool. SME Rapid Prototyping and Manufacturing Conference Proceedings, 1996. F Prioleau. Comparison of SLS RapidTool process to others. SME
Rapid Prototyping and Manufacturing Conference Proceedings, 1996. E Sachs.
Injection molding tooling by three dimensional printing. SME Rapid Prototyping
and Manufacturing Conference Proceedings, 1996.
2. KR Denton. Quick Cast and rapid tooling: A case history at Ford Motor Company.
SME Rapid Prototyping and Manufacturing Conference Proceedings, 1994. R Erikson. Cast tool prototyping for injection molding: Where is it going? Sixth International Conference on Emerging Technologies and Business Trends in Plastics
Injection Molding, March 1996.
3. LE Weiss, EL Gursoz, FB Prinz, PS Fussel, S Mahalingam, EP Patrick. A rapid
manufacturing system based on stereolithography and thermal spraying. Manuf
144
4.
5.
6.
7.
8.
Wise
Rev 3(1):40, 1990. K Maley. Using stereolithography to produce production injection molds. ANTEC 94, p. 3568. C Hefright. Applying laser technology to rapid
prototyping. ANTEC 93, p. 406. A Mathews. Nickel vapor deposition tooling for
the plastics industry. Proceedings of the Third International Conference on Advances in Polymer Processing, March 1993.
Modern Plastics Encyclopedia 94. New York McGraw-Hill, 1994, p. 185.
Modern Plastics Encyclopedia 94. New York: McGraw-Hill, 1994.
L Miller, S Wise. Chemically bonded ceramic tooling for advanced composites.
Mater Manuf Process 5(2):229252, 1990.
S Wise. Net shape nickel ceramic composite tooling from RP models, SME Rapid
Prototyping and Manufacturing Conference Proceedings, 1996.
DA Van Putte, LE Andre. A step-by-step evaluation of building an investment cast
plastic injection mold. SME Rapid Prototyping and Manufacturing Conference
Proceedings, 1995.
7
Nickel Vapor Deposition Technology
Debbie Davy
Mirotech, Inc.
Toronto, Ontario, Canada
Ni 4CO Ni(CO)4
These molds can be used in conjunction with rapid-prototyping patterns in
many diverse applications, such as injection molding, blow molding, net nickel
shapes, and so forth. This technology allows mold-makers, molders, original
equipment manufacturers, and others to respond to rapidly changing markets,
reducing costs and shortening product-development time. The NVD process
offers the potential for rapid turnaround times, accuracy on an atomic level,
and faster cycle times relative to existing manufacturing processes.
Figure 1 is a schematic of the NVD process, utilizing nickel carbonyl
gas, Ni(CO)4. This gas provides a useful property: Specically, it breaks down
into a solid metal at a high rate of decomposition. By heating a mandrel or
substrate to the required temperature and having Ni(CO)4 gas ow over the
mandrel in a sealed chamber, an exact nickel negative of the mandrel is obtained. Deposition rates from 50 to 750 m/h (0.0100.030 in./h) can be obtained on surfaces held at temperatures between 110C and 190C. The nickel
146
Davy
II.
Over the past 10 years, all major global manufacturers have experienced tremendous competitive pressures. The Japanese and Pacic Rim countries have
utilized their low-cost, high-quality production techniques to gain a signicant
level of penetration into both North American and European markets. These
competitive pressures are forcing North American and European manufacturers to look for new innovative technologies to help them in reducing the
price of their product and especially in reducing their product-development
time.
There is, however, a major problem associated with traditional moldmaking techniques which results in signicant delays from the approval of
the part design to the completion of the nished tool. This is the time required
to machine the core and cavity inserts of a mold to the precise dimensions
required, typically resulting in mold delivery times from 16 to 30 weeks.
147
148
IV.
Davy
5.
6.
7.
8.
9.
10.
11.
12.
Shapes/cavities are created with uniform wall thickness throughout. External corners build up at the same rate as at surfaces.
Internal corners do not show a signicant reduction in deposit
thickness, resulting in a uniformly strong tool face.
Deposition can be achieved on appropriate substrates without affecting the surface delity of the master.
Nickel provides a mold surface with high-temperature and corrosionresistant properties suitable for molding aggressive resins such as
phenolics or vinyls.
Nickel offers excellent release properties.
Nickel vapor deposition produces parts with low levels of internal
stress. This results in minimal warping of the nished shells, better
matching cores, and cavities.
The NVD process is generally less expensive than traditional tooling, offers fast turnaround time and rapid deposition (e.g., a 12.7mm-thick NVD nickel shell can be generated in less than 2 days,
regardless of the tool size). The NVD process typically deposits
nickel at rates from 50 to 750 m/h (0.0100.030 in./h).
The surface of the master is replicated with outstanding surface
denition; for example, optical quality surfaces have been replicated.
Uniform shell wall thickness is produced, even around acute
angles. The NVD process relies on the thermal decomposition of
nickel onto the substrate and not an electrolytic deposition. This
means that if the surface temperature of the substrate/mandrel is
uniform, the thickness of the deposit will also be uniform! This is
true regardless of the geometry or shape of the substrate surface.
Sharp internal corners can be used, resulting in increased design
freedom.
The NVD nickel molds can be repaired by TIG welding or brazing.
The NVD nickel is more dense and ductile than electroformed
nickel. This results in a tougher and stronger tool face capable
of withstanding additional repeated tooling process cycles and a
reduction in the tendency for the nickel tool face to crack in use.
The NVD nickel has very little porosity. This provides the following benets: (a) NVD nickel polishes extremely well; (b) there is
no fear of breaking through the surface into porous areas of nickel;
and (c) NVD nickel contains almost no sulfur whatsoever and can
13.
149
150
Davy
VI.
NVD APPLICATIONS
The NVD process has been used in many diverse applications, such as the
following:
1.
2.
3.
4.
5.
6.
7.
VII.
The work on nickel that has been deposited by CVD has yielded data that
falls into several categories. These are tensile properties, hardness properties,
thermal expansion, residual stresses, chemical composition, and microstructure. These data have been obtained from tests and experiments that have been
performed at Ortech, at the University of Toronto, at Ontario Hydros Research Division, and at Camnet.
Coefcient of thermal expansion: 13.1 106 m/m/C
Thermal diffusivity: 3.66 m2 /h
Thermal conductivity: 88 W/mC
Residual stress (surface): 3060 MPa tensile
Yield strength: 584 39 MPa
Ultimate tensile stress: 827 7 MPa
Modulus elasticity: 157224 GPa
Elongation: 612.4%
Hardness: 1550 RC (variable)
Chemical composition: Ni
C
S
H
151
99.98%
150 ppm
1.0 ppm
7.3 ppm
152
Davy
VIII.
Both NVD tooling and electroformed tooling can offer substantial cost-saving
benets over machined steel tooling, especially for larger parts. Also, a substantial time savings can often be realized relating to shorter tool-generation
times. Electroformed nickel, although also capable of generating production
tooling, has some inherent limitations that must be considered before a successful tool can be designed.
The NVD process relies on the thermal deposition of nickel onto the
substrate, not an electrolytic deposition. This means that provided the surface
temperature of the substrate/mandrel is uniform, the thickness of the deposit
will also be uniform. This is true regardless of the geometry or shape of the
substrate surface. With electroformed nickel tooling, surface geometry has a
marked effect on deposit thickness. External corners on the mandrel exhibit
higher local electric elds and will show an increased buildup of nickel. Conversely, internal corners result in lower local electric elds and will show a
decrease in deposit buildup, as seen in Fig. 4A. Very sharp internal corners
will accentuate this effect. The result can often be an internal corner with 50%
or less nickel deposit than the average thickness. The thinner nickel shell will
be inherently weak, which can result in premature failure of the tool in this
location. This inherent electroforming limitation often results in a redesign of
the mandrel to minimize shell thickness variations.
(A)
153
(B)
With the NVD process, external corners build up at the same rate as the
balance of the surface, as shown in Fig. 4B. The internal corners do not show
a signicant reduction in deposit thickness, resulting in a much more uniform
shell thickness and, hence, more uniform shell strength. Very sharp internal
corners can be used, providing increased design freedom. With electroforming,
this is not always possible.
The NVD nickel is considerably denser and much more ductile than
electroformed nickel. This results in a tougher and stronger tool face capable
of withstanding a greater number of repeated tooling process cycles, with a
marked reduction in any tendency for the nickel tool face to crack in use.
Electroformed nickel often contains both surface and internal porosity. NVD
nickel has very little porosity.
The NVD nickel polishes better than electroformed nickel, and there is
no fear of breaking through the surface into porous regions of nickel. Electroformed nickel can contain sulfur, which results in poor welding performance.
NVD nickel contains almost no sulfur whatsoever and can be welded readily.
This means that repairs or modications to the NVD tool surface can be performed at any time.
Both electroformed nickel and NVD nickel can provide tool faces with
a hard surface for abrasion resistance. However, when electroformed nickel
is produced with a hard face, the process generally requires two steps:
1.
2.
154
Davy
A defect that can occur due to this two-step process is delamination. If the
activation of the rear of the nickel face is not 100% successful, then stress
concentrations can occur at the interface between the two different hardness
layers. Potential delamination is difcult to inspect and may occur much later
during tool operation.
IX.
The majority of large molds are made by CNC and EDM machining out of
steel. Some tools intended for limited production are also made from aluminum, which is less expensive to machine than steel. For most low-pressure
plastic injection-molding applications, steel tooling is overengineering and involves an unnecessary expense. Additionally, many large plastic parts are
lled with glass bers. Hence, abrasion is an important issue to consider in
mold design. Consequently, steel often supercedes aluminum as the material
of choice. Here, an NVD composite mold can have an enhanced surface hardness to ensure a long-lasting tool surface. A softer substrate such as copper
or aluminum-faced epoxy can become a viable alternative to steel molds for
moderate to large production runs.
Steel molds are very heavy compared to NVD composite molds. Consequently, handling and storage is an important consideration. Design, fabrication, assembly, checkout, and delivery of a steel mold can take 1626 weeks
or more, whereas an NVD composite mold can be completed in less than half
that time.
X.
XI.
155
NVD nickel can be deposited onto any surface provided the following hold:
1.
2.
3.
Unfortunately, this is not the case for mandrels made from the following:
Stereolithography
ACTUA-2100
SLS wax
The Saunders process
The Cubital solid-ground curing process
Fused deposition modeled (FDM) waxes (ICW04 and MW01)
156
Davy
The backing has its CTE matched to that of the NVD shell.
The backing is well bonded to the NVD nickel shell.
The ejector-pin and sprue bushings are encapsulated in NVD nickel.
The steel frame is also encapsulated in NVD nickel.
The backing material is thermally conductive.
157
Figure 6 Cross-section of an NVD mold. (1) NVD nickel face; (2) outer support
frame of fabricated steel with encapsulation; (3) heating/cooling coils, usually copper
located at rear of shell; (4) ejector-pin bushings and gates (note encapsulation); (5)
cast backing (composite with matched CTE); (6) bonding aides (stud welding).
158
Davy
XII.
Tooling for the plastics industry is typically produced in two major forms:
1.
2.
Production Tooling: machined steel; machined aluminum; electroformed nickel shell with cast resin backing
Prototype Bridge Tooling: cast epoxy or ber-reinforced composites; spray metal shell with cast backing
159
8
The ExpressTool Process
Paul F. Jacobs
Laser FareAdvanced Technology Group
Warwick, Rhode Island
I. INTRODUCTION
Previous chapters of this book have described the business signicance of
rapid time-to-market and the potential impact of reduced tooling lead time.
Also discussed were techniques for the generation of rapid soft tooling as
well as recent advances in rapid bridge tooling. Additional chapters describe various approaches to rapid production tooling, including their advantages and benets, as well as their shortcomings. In this chapter, we shall
describe the ExpressTool process in some detail.
ExpressTool evolved from a joint project between the Hasbro Corporation (Pawtucket, RI) and Laser Fare, Inc. (Smitheld, RI). As one of the
worlds largest toy companies, Hasbro generates a great number of plastic
injection molds every year. Aware of the importance of rapid time-to-market
in the highly competitive toy business, Hasbro formed a strategic partnership
with Laser Fare in 1992. During this collaboration, a number of different rapid
tooling approaches were investigated. The most successful was based on the
electroforming work performed by Richard Barlik, and is the predecessor of
the ExpressTool process. Although numerous modications have been made
since, the basic physics and chemistry of the patent pending process were
developed jointly by Hasbro Corp. and Laser Fare from 1992 to the present.
When the ExpressTool process had achieved an appropriate level of
maturity and repeatability, an extended beta test program was started. This
162
Jacobs
led to the fabrication, assembly, and operation of molds for various corporations that were well aware of the importance of rapid time to market. This
test program continued throughout 1997. In March 1997, ExpressTool, Inc.
was formed as a wholly owned subsidiary of Innite Group, which also owns
Laser Fare. The ExpressTool process was ofcially commercialized as of January 1998, for the purpose of fabricating high-productivity production tooling.
Currently, production molds are being built for automotive, aerospace, and
consumer product manufacturers.
II.
HIGH-THERMAL-CONDUCTIVITY MATERIALS
Figure 1 plots thermal conductivity for some relevant mold materials. Heat
transferred from the plastic must be conducted through the mold before it can
be removed by coolant. Thus, the thermal conductivity of the mold directly
impacts the speed of the injection-molding process. Inspection of Fig. 8 reveals
one of the basic problems with steel, by far the most common material used
in building production molds.
Here, H-13 tool steel, having a thermal conductivity of 28 W/m K was
chosen to be representative of the broad class of tool steels. As a point of
interest, 316 stainless steel is even less thermally conductive, at about 20 W/
163
164
Jacobs
ment, behaves like a classic bimetallic strip. The result is substantial internal
stress and plastic part distortion!
Thus, an important if not often articulated goal in plastic injection molding is to improve the uniformity of the active mold surface temperature distribution over time. Finite-element analysis (FEA) results presented in this chapter show that conformal cooling channels (CCCs), in conjunction with high
conductivity mold materials, can provide substantial temperature uniformity
benets. By optimally positioning the CCC in x, y, and z space, it is possible
to further reduce mold temperature variance.
A key measure of mold performance is Tmax, dened as the difference
between the highest temperature of the active mold surface and the lowest
temperature of the active mold surface at the instant the rst sector of plastic
begins to solidify and shrink. Lower values of Tmax provide more uniform
shrinkage and, consequently, less part distortion (13).
Figure 2 shows a conformal cooling channel used in the injection molding of a Vaseline jar cap for Chesebrough-Ponds. Note that the CCC transitions
from a straight vertical section into an oval shape in the horizontal plane and
back to vertical again. Machining a channel of this geometry in a solid block
of steel would be impossible in a single piece, or prohibitively complex and
expensive in multiple sections. However, when the active surface of the tool
has been electroformed as a thin nickel shell, then positioning CCC behind
that shell becomes relatively straightforward.
165
IV.
The key aspects of the ExpressTool process are described below. The procedure can best be illustrated by following a typical mold-fabrication sequence:
1.
2.
166
Jacobs
with reduced shrinkage and improved green strength (6), and (c)
the deployment of advanced build techniques (7). Nonetheless, with
the possible exception of recent accuracy results by Sanders (which
are not yet statistically complete), none of the RP&M processes can
match CNC accuracy and repeatability for lengths beyond about 4
in. For dimensions over 10 in., RP&M mandrel errors often exceed
three times the corresponding CNC values.
Further, a number of commercially available, CNC-machined
materials currently provide better dimensional stability than any
RP&M-generated parts. This is true for ambient conditions (i.e.,
sitting on a bench, waiting for the next step), and also for electroforming conditions (i.e., being immersed in a vat containing a warm,
aqueous, electroforming solution).
Figure 3 shows a CAD model of a mandrel used as a test
sample to establish dimensional stability. Initially, this CAD test
part was CNC machined in the test material. Next, a number of
important dimensional characteristics (e.g., the atness of both the
top and bottom planes, the straightness of the vertical walls, and
the values of both the interior and exterior dimensions in the X,
167
168
Jacobs
3.
169
170
Jacobs
nite-element analysis (T-FEA). The T-FEA would utilize the original 3D solid CAD model to establish that CCC path which results
in the minimum active mold surface temperature difference, Tmax.
The major benets of reducing Tmax are as follows:
Reduced cycle time. One must wait until the last portion of
the injected plastic cools below its heat deection temperature
before ejecting the part. Minimizing Tmax reduces this wait,
decreases cycle time, and increases overall productivity!
Reduced part distortion. Nonuniform active mold surface
temperature distributions result in variable part cooling rates
and sector-delayed shrinkage, leading to greater part distortion.
7.
171
8.
9.
Signicant differential expansion and contraction could lead to substantial induced stress, plastic deformation, fatigue, delamination,
and failure of the tool. At this time, NiCu/CCC inserts have already achieved 270,000 shots with no delamination problems. Data
for a substantially greater number of injection cycles are currently
being developed as a result of an exclusive agreement for joint testing between the General Electric Plastics Division (Pittseld, MA)
and ExpressTool (Warwick, RI).
The combined mandrel/nickel shell/copper thermal management
layer ensemble, including the encapsulated conformal cooling
channel, is then backed with an insulating material. Once the heat
has been successfully transferred to the cooling channels, there is
no point in providing high-thermal-conductivity material further
into the tool. Thus, a number of commercially available lled epoxy
formulations can provide good compression strength with relatively
rapid cure (24 h). When mold pressures are expected to exceed
10,000 psi, machined steel is used.
Next, the core and cavity inserts are positioned in a mold frame.
The ejector holes are now machined through the backing layer as
well as the copper thermal management layer and the nickel shell.
Of course, care must be taken to avoid drilling through a conformal
cooling channel. In this regard, an additional advantage of T-FEA
positioned conformal cooling channels lies in the accurate location
of the channels as they appear in the solid CAD model. The CAD
model can then be used as an aid in locating ejection pins to ensure
172
Jacobs
V.
CASE STUDY 1
173
VI.
CASE STUDY 2
Figure 6 shows a CAD model of a standard Vaseline jar cap injection molded
in high-impact styrene for Chesebrough-Ponds. The performance of an existing H-13 steel mold built with conventional DCC was compared to the
performance of an electroformed NiCu tool with encapsulated CCC.
After setup, thermal stabilization of the tool, and optimization of the
molding parameters, the measured cycle time for the production H-13/DCC
mold was 15 s, corresponding to 3600/15 240 parts per hour, assuming
uninterrupted operation of the molding press. Again, after setup, thermal stabilization of the tool, and optimizing mold parameters, the cycle time for the
electroformed NiCu/CCC mold was 9 s, corresponding to 3600/9 400
parts per hour, assuming uninterrupted operation of the molding press. Note
that 400/240 1.67, or a 67% increase in mold productivity when using
electroformed nickelcopper core and cavity inserts with encapsulated conformal cooling channels!
It is clear from these two case studies that the reduction in mold cycle
time and the consequent increase in productivity for NiCu/CCC molds relative to conventional H-13/DCC steel molds is dramatic. In Sec. VII, FEA
results provide an explanation for these substantial reductions in cycle time,
as well as major improvements in mold temperature uniformity.
174
VII.
Jacobs
FINITE-ELEMENT ANALYSIS
175
CCC geometry looks something like a keyhole when viewed from above.
Although, in principle, the CCC could also have arbitrary cross-sectional
shape, the channel cross sections were assumed to be circular for this study.
In future studies, we will evaluate the effects of noncircular channel cross
sections.
Figure 8 shows the model of the NiCu tool developed at NRC by Michel Perrault, which formed the basis of the ensuing FEA analysis. The following assumptions were made:
176
Jacobs
The difference in the two temperature distributions is dramatic! The H13 tool with DCC shows a hot spot to the left of the cooling channel (near
the sprue) and another to the right of the channel. Conversely, the NiCu tool
with CCC shows an almost isothermal temperature distribution. The value of
Tmax for the H-13/DCC case is 12.5C. In contrast, the value of Tmax for the
nickelcopper tool with CCC is only 2C. Obviously, the combination of
high-thermal-conductivity materials and conformal cooling channels has
signicantly reduced mold temperature variations in this case.
Figure 10 is another FEA image, this time of the temperature distribution
on the active mold surface of the cavity side of the tool for the conventional
H-13 tool with DCC on the left and the NiCu tool with CCC on the right.
At the active mold surface the effect is even more dramatic. The value
of Tmax for the H-13/DCC cavity is 18.6C, and the corresponding value for
the NiCu/CCC cavity is only 1.9C, or, essentially, an order of magnitude
reduction in active mold surface temperature variance!
Figure 11 shows the pseudo-color temperature distribution for the cavity
surface of the H-13/DCC tool at 2-s intervals from 1 to 15 s after plastic
injection. These images illustrate the cooling of the insert over time. Figure
12 shows the same information for the NiCu/CCC tool. It is clearly evident
from inspection of these two gures that the cooling rates for the NiCu/CCC
tool are much faster than for the H-13/DCC tool. In fact, the temperatures
177
178
Jacobs
Figure 12
throughout the NiCu/CCC tool only 3 s after injection are already lower
than the corresponding temperatures for the H-13/DCC tool after 15 s!
These data begin to explain the reasons behind the extraordinary productivity improvements noted in the two case studies presented in Sects. V and
VI. The only reason the productivity gains are not even greater is that the
cycle time includes not only the cooling time but also the times needed to (a)
close the press, (b) inject the plastic, (c) pack the plastic, (d) open the mold,
and, nally, (e) eject the part. However, neither the thermal conductivity of
the mold nor the presence of CCC has any effect on these ve time intervals.
Thus, the dramatic productivity gains documented for NiCu/CCC inserts are
purely the result of signicantly reducing the mold cooling time.
VIII.
PROCESS CHARACTERISTICS
Electroformed nickelcopper inserts with encapsulated CCCs provide a number of important benets, as well as some limitations. These are discussed in
this section.
1.
Thermal conductivity. As noted previously, the thermal-conductivity values of nickel and copper are both dramatically higher than
any of the various tool steels. Assuming 2 mm electroformed nickel
2.
3.
4.
5.
6.
179
180
Jacobs
7.
8.
9.
10.
all conventional tool steels. The best of the conventional mold materials used when injection molding reactive plastics [e.g., poly
(vinyl chloride)] are stainless steels. Indeed, nickel is used as an
alloy ingredient in stainless steel to improve chemical resistance.
Experience has shown that NiCu/CCC inserts exhibit virtually
no signs of chemical attack during the injection-molding process.
Surface quality. Electroformed nickel surfaces can be highly polished and have been used for many years in the injection molding
of plastic eyeglass lenses. Optical quality surface nishes as good
as Ra 2 in. (0.05 m) have been routinely achieved on electroformed nickel.
Textured surfaces. Mold-Tech, Inc. has successfully textured the
active electroformed nickel surfaces of NiCu/CCC inserts. According to Mold-Tech, the resultant texturing using their standard
procedures was sharp, well dened, and capable of good depth
when needed.
Mold repair. A truly unique aspect of building production tools
through the use of the electroforming process is the capability to
reelectroform. In the event that a glass-lled plastic has gradually eroded any portion of the active surface of the tool, it is possible to simply mask the unworn portions of the insert and then
reelectroform the worn surface. Because the nickel electroforming
process adds material at about 1 m every 5 min, it is possible to
rebuild worn areas in a very controlled manner. Obviously, if the
tool surface is textured, the rebuilt area will also require subsequent texturing. Of course, the same would be true for a conventional steel tool that had undergone weld repair. One major difference, however, is that weld repair involves considerable heat input
and the possibility of insert distortion. Conversely, electroforming
is performed in a warm bath involving negligible heat loading and
essentially zero insert distortion.
Size. The electroforming process is not fundamentally or intrinsically limited in size by any accuracy, plating, or processing step.
Because CNC is certainly capable of producing large mandrels
accurately, and electroforming involves essentially zero randomnoise shrinkage, the only limit at present involves the size of the
vats. The current ExpressTool electroforming vats are about 3 ft
wide by 5 ft long by 2 ft deep. This has been more than sufcient
for all projects performed to date. Should larger inserts be required,
181
2.
182
Jacobs
3.
ACKNOWLEDGMENTS
The author would like to acknowledge the outstanding cooperation of the Process Modelling and Optimization group at the the National Research Council,
Boucherville, Quebec, Canada, under the direction of Georges Salloum, and
especially the extraordinarily capable and creative efforts of Michel Perrault.
Mr. Perrault developed the Finite Element Analysis model for the conventional
H-13 core and cavity inserts with drilled cooling channels, as well as the FEA
model for the electroformed nickelcopper core and cavity inserts with encapsulated conformal cooling channels. The FEA temperature distributions presented in this chapter were the result of his excellent efforts.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
183
tooling in plastic molding. Report No. 397, Department of Mechanical Engineering, Ohio State University October 1991.
B Bedal, H Nguyen. Advances in part accuracy. In: P Jacobs, tech. ed. Stereolithography and Other RP&M Technologies. Detroit, MI: SME New York: ASME,
1996, pp. 164180.
P Jacobs. Fundamental processes. In: Rapid Prototyping & Manufacturing: Fundamentals of Stereolithography. New York: SME/McGraw-Hill, 1992, pp. 79
110.
T Pang. Advances in sterolithography photopolymer systems. In: P Jacobs, tech,
ed. Stereolithography and other RP&M Technologies. Detroit, MI: SME/New
York: ASME, 1996, pp. 2779.
B Bedal, H Nguyen. In: P Jacobs, tech. ed. Detroit, MI: SME/New York: ASME,
1996, pp. 156162.
T Mueller. A model to predict tolerances in parts molded in pattern based alternative tooling. Proceedings of the 1998 SME Rapid Prototyping & Manufacturing
Conference, Dearborn, MI, May 1998, pp. 559577.
K Filipiak. Injection molding thermoplastic parts in days in tooling produced
from new composite board. Proc. 1998 SME Rapid Prototyping & Manufacturing
Conference, Dearborn, MI, May 1998, pp. 223243.
International Nickel Co. Electroforming with Nickel. American Electroplaters
Society, Inc., 1997, pp. 1214.
9
An Automotive Perspective
to Rapid Tooling
Anthony T. Anderson
Ford Motor Company
Redford, Michigan
I. INTRODUCTION
On a global and domestic scale, Americas share of the automotive market
has decreased primarily due to increasing foreign competition and rapid market growth in Asia and South America. In 1965, U.S.-based manufacturers
produced over 53% of all vehicles sold in the world, with an 8% average
return on sales. Today, the United States makes only 36% of all vehicles sold
in the world with less than 2.5% return on sales (Automotive Industries, November 1997, p. 5). Since the end of the cold war, the U.S. automotive industry
has been forced to change to become more competitive in a rapidly growing
global economy. The industry has pushed to institutionalize processes that
provide speed to the marketplace: simultaneous engineering, agile manufacturing, world-class timing, and corporate globalization. The Japanese have provided the benchmarks for change, where quality and cost competitiveness have
become required entry fees to the game. These competitive challenges have
put a strain on U.S. automotive manufacturers to maintain their share of the
market with a production system that evolved in the absence of these concerns.
In response, efforts are being made to incorporate processes that improve communication both internally and within the supplier base, to take full advantage
of our diverse workforce and become more exible as the market continues
to become more global. These efforts provide a basis for recognizing potential
rapid tooling (RT) applications from an automotive perspective.
186
Anderson
Current trends to reduce the product-development cycle time and manufacturing cost in the automotive industry are discussed in terms of (a) our
utilization of rapid prototyping and manufacturing (RP&M) to accelerate the
product design process and (b) the emergence of rapid tooling (RT) technologies for future low-cost niche market manufacturing. These trends involve the
integration of computer-aided fabrication technologies with proven low-cost
fabrication processes to develop more economical manufacturing methodologies with improved system robustness. From an automotive manufacturing
perspective, successful implementation will rely on our industrys ability to
improve communication through cross-functional team efforts while reducing
technology development costs through multiple-resource leveraging. Meeting
these challenges of change will be key to survival for the North American
Automotive Industry in the 21st century.
A.
187
Even if each machining operation can be done quickly (high-speed machining), setup time and idle time periods between each stage can be extensive.
Efforts have been made to reduce these bottlenecks between machining operations by utilizing ve-axis machine cells that combines operations (agile manufacturing). These machining systems are more exible than dedicated machines for part manufacture (2,3). Unfortunately, their associated high variable
cost and complex tool path generation make implementation for competitive
low-volume manufacturing difcult to justify. The existing problem is that
whereas computer-aided design (CAD) can easily design complex parts and
todays machine tools can easily and efciently cut them (high-speed machining), the process by which the multiaxis machining motion is described has
not changed signicantly for almost 30 years. Despite reports that automated
tool path generation has made signicant progress reducing productdevelopment lead times (4), Automatically Programmed Tool (APT), the underlying mathematical technology for multiaxis machining, does not meet todays machining needs. Except for specic cases where parts have smooth
contoured surfaces (stamping dies), the highly skilled APT programmer must
discretely program every surface and check for each potential gouge, tangent,
or surface discontinuity. Although APT-based systems can program complex
parts, these systems take long times to learn and the programs generated are
characteristically complex and are difcult to verify. Nelson Metal Products
claims to have developed a time-saving software program capable of generating tool path data directly from CAD data with minimal human intervention. These tool path data are used to make a complex prototype part by computer numerically controlled (CNC) machining. Nelson uses this RP&M
tooling fabrication approach to reduce the lead time for optimizing Fords
Front End Accessory Drive bracket design. Unfortunately, their approach is
limited to this specic application. To date, truly automated tool path generation software is being successfully applied only to specic applications. A
more general CNC software program has yet to be developed that can generate
tool paths directly for ve-axis machining of arbitrarily complex parts without
some highly specialized, human intervention. In general, NC programming
with its associated high level of human interaction remains the major bottleneck in the product-development process.
To date, low-volume product developments have been fueled by a successful systems approach to lowering fabrication cycle time among OEMs in
the auto industry. For example, assemblies like Fords Sheet Molded Composite Aerostar hood are now produced as a single part in a minute or less. Although consolidating this assembly into a single more complex part takes
longer, the total fabrication time and cost is far less than what is required to
188
Anderson
form and join simpler designed components together. Likewise, General Motors thermoplastic (30% glass-lled polycarbonate/polyester) door module
consolidates as many as 61 individual parts (most were metal stampings),
reducing assembly time by 84% on their minivans, Chevrolet Malibu, and
Oldsmobile Cutlass doors. DaimlerChryslers Composite Concept Vehicle,
once known as the China Concept Vehicle, represents the ultimate in part
consolidation. The entire body shell consists of just four injection-molded
composite plastic components (15% glass-lled polyethylene terephthalate).
These 4 components would replace over 80 stamped and welded parts in a
typical steel car body. Also, ferritic stainless exhaust manifolds (currently a
stamped, tubular, welded assembly) can be hydroformed with 33% fewer operations and 20% fewer assembly components. In addition, injection-mold and
hydroform tooling requirements are much less severe than those for traditionally stamped components, making their use ideal for low-volume fabrication.
Other cost-saving fabrication technologies for low-volume component manufacture include reaction injection molding (RIM) and resin transfer molding
(RTM). These specialized processes help lower overall cost and vehicle weight
for specic part applications such as structural components (underbody crossmembers, oor pans, and other body parts) by replacing traditional steel
stampings with lighter weight, ber-reinforced, plastic composites. Although
composite parts take longer to make than steel stampings, tool requirements
for RIM and RTM parts are much less severe. As a result, urethane or Nishell molds with cast aluminum, epoxy, or cement backing can be used for
production of composite structural components at a fraction of the cost and
lead time of traditionally machined tool steels (5,6). The utilization of shellmold designs for accelerated tool fabrication is reviewed in Sec. V.
Another approach to being competitive in the niche-car market is to go
global. High-speed communication technologies allowed Ford leadership to
develop a genuine global car, the CDW27. When Ford of Europe needed a
new mid-sized family car, with a market potential of only 25,000 units a year,
the company could not make a prot building a sophisticated niche vehicle
for one region. However, by spreading the cost of development and production
around the world, a 100,000 units per year market potential could be realized,
where the high initial cost could be offset. This strategy allowed Fords
CDW27 to become the rst modern global car. It was named Mondeo in Europe, Taiwan, and the Middle East, and slightly modied versions went on
sale in North America with the names Ford Contour and Mercury Mystique.
The success of Fords CDW27 became the prototype for a new way of thinking about a range of product developments with common platforms. It proved
that true globalization was nally possible with enabling communication tech-
189
nologies and that customer-focused teams were the way of the competitive
future.
Unfortunately, these efforts to lower overall product-development cost
are effective for specic applications only. They still lack the robustness
needed to allow us to be more competitive in the niche-car market in general.
Some insight to this problem can be made more apparent by reviewing our
traditional product-development process and how computer-aided technologies can assist.
Figure 1 Product-development time lines A: concept development; B: prototype development; C: manufacturing development. **Development time minimized to show
potential of rapid prototype (RP) utilization. (Data from Automotive Industries, September 1991.)
190
Anderson
competition was localized within the country. These companies grew by being
less cooperative and more competitive. The traditional methods of communication between product and manufacturing engineers became the infamous
toss it over the wall approach. There was little collaboration in the early
stages of the product-development cycle. As the automotive market became
more global, customer demands for sophisticated niche cars grew to meet
their ever-changing social and environmental expectations while government
regulations increased for cleaner air and greater fuel economy. These changes
increased vehicle manufacturing and organizational complexities both internally and within the supplier base. Unfortunately, the result was longer
product-development lead times and higher product cost.
This situation is changing rapidly among U.S. OEMs in the automotive
industry. Conventional thinking, limited to the type of machines and methods
used in the past, is giving way to radically new approaches to reducing productdevelopment times. Figure 1 forecasts how the integration of RP&M into
the product development process can reduce overall cycle time by over 50%,
making a U.S. OEM more competitive than ever. This forecast is based on
the accumulated inuences of rapid prototyping (RP) on the prototypedevelopment stage and RT on the manufacturing-development stage of the
product-development cycle. In general, the walls of communication between
product and manufacturing are being broken down in the United States by the
use of computer-aided technologies.
Figure 2 is a simplied model representing the industrys major communication stages of product development from concept to customer: (1) concept
design, (2) prototype verication, (3) tooling fabrication, (4) manufacturing
process feasibility, (5) assembly optimization, and (6) customer approval. Traditionally, product-development communication only ows downstream from
concept to the customer. When one stage of the process is completed, information is tossed over the wall to the next stage. This one-way approach
to information ow is characterized by many costly, time-consuming, engineering changes that occur further downstream, making cost-effective
globalization difcult to achieve. In contrast, upstream communication
ow, like customer-driven concept developments (listen to the voice of the
customer), helps improve sales, and predicts future markets. Likewise,
computer-aided technologies like Design for Assembly (DFA) and Design for
Manufacturing (DFM) help improve product quality and reduce manufacturing
cost. Additional upstream information ow between the manufacturing
process and tooling fabrication stages encourages process-driven tool development for reduced fabrication lead time and cost (a rapidly growing future
trend).
191
192
Anderson
C.
193
194
Anderson
(CAD/LAM), prole edge lamination (PEL), and direct slice control machining (SCM).
A.
To make functional prototypes from engineering metal materials, this (computeraided) slice or layering approach has been successfully applied to traditional
three-axis CNC machining. This exible processcalled precision stratiform
machiningdivides the part into layers (of varying thicknesses depending on
the part complexity), generates three-axis tool paths for each layer, and machines each layer to dimension from the desired wrought engineering material
stock. This layering approach signicantly reduces user interface time for tool
path generation, making the process more robust than traditional CNC machining of complex parts. The layers or plates are subsequently stacked horizontally and vacuum brazed together. One of the rst components developed using
precision stratiform machining was a prototype aluminum cylinder head for
Fords 2-L Zetec engine (14). This part is too complex to be made directly
by conventional NC machining from wrought stock. A six-layer working prototype of this complex engine component was made using the precision stratiform machining process in just 100 days, a third of the time required to make
traditionally cast cylinder-head prototypes.
B.
CAD/LAM
The CAD/LAM process incorporates the integration of CAD with CAM laser
cutting, adhesive bonding, brazing, and mechanical fastening to construct laminated steel molds. This layering approach to traditional machining can also
be used to make laminated steel or aluminum tools in a fraction of the time
required for traditional machining from wrought or cast metal stock. Fords
Climate Control Division and Toledo Mold and Die have jointly investigated
the process for making injection molds. Like the precision stratiform machining process, CAD/LAM utilizes a 3D solid model to dene both the part surface outline as well as component details contained within each individual
layer. Data for each section are then translated into CNC cutter paths which
will be used to cut the outline and internal features of each layer to form stock
material. Unlike the stratiform process, CAD/LAM uses sheet stock material
of constant thickness. Each sheet lamination is thin enough to be easily machined to prole using a laser or traditional CNC machine. The sheets are
then stacked laying horizontally, one on top of another, mechanically attached
with vertical fasteners, and brazed together by capillary ow of liquid copper
195
between each plate. The steps formed on the molds surface by the sharp edges
of the stacked layers are removed by electrical discharge machining (EDM).
The CAD/LAM process offers a potential material cost and time savings
when compared to traditional tools machined from wrought stock. In addition,
this laminated-die fabrication approach offers improved cooling capabilities
for injection-molded tools. Water lines can be routed to follow, more closely,
the contour of the molds surface geometry and achieve a more uniform cooling of the part (i.e., conformal cooling). Unfortunately, horizontally stacked
sheets are difcult to align and secure to one another because clamping must
be done through all the sheets to hold them together properly. Also, when
thinner sheet stock is used, it has less resistance to warpage during the brazing
process. Efforts have been made to replace the brazing process with an adhesive bonding process with some degree of success (15). However, bond
strength between layers were found to be directly related to how uniformly
the adhesive can be distributed. Using current bonding techniques, it is difcult
to maintain an even spread of adhesive with no voids. Further developments
are needed in this area before production applications can be addressed for
the CAD/LAM process. In addition, the greater the number of sheets used,
the greater the problem clamping and brazing the assembly together. This
stacked edge bonding problem can be overcome by changing the orientation
of the stacked sheets as is done in the PEL process.
196
Anderson
gether remain a problem that must be fully resolved before use of these processes become widespread.
D.
This process being developed at Clemson University overcomes the laminatebonding issue by machining a layer at a time from a solid block of material.
This approach utilizes the same robust slice control building algorithms characteristic in most RP machines. The SCM process incorporates the use of a
computer hardware/software interface for converting CAD slice les directly
into NC code machining data with little to no user interaction. Even though this
approach eliminates problems associated with holding the laminated sheets
together, the material removal process unfortunately lacks the potential for
fabricating improved internal features such as conformed cooling channels in
injection molds. Unlike traditional machining, however, the SCM approach
simplies the tool path generation process, giving it the potential of economically machining accurate, quality, prototype patterns and tools faster than current CNC machine technology.
197
A. Concept Models
Touch/feel prototypes or concept models are commonly used to communicate design concepts, verify geometric shape intent, and to check some t
issues during assembly with other parts in the early stages of the productdesign cycle. Traditionally, model shops work from 2D part drawings to either
machine a master part directly or make an original model form from clay
or other soft sculpting material like wood or foam. Design errors are noted
and a new or modied concept model is made. This procedure is repeated
over and over until a visual design intent is veried. Accuracy is not a
critical requirement for concept models. For small parts, RP models (with
tolerances of 0.003 in.) have been successfully used for this application,
typically reducing lead time from 8 weeks to 3 weeks. Because RP&M machines work from 3D CAD data instead of 2D drawing data, design misinterpretations are eliminated and undetected human errors are minimized. To date,
part size has been limited to the envelop size of the RP&M machine. Larger
part models have been made by the assembly of smaller pieces, but timesaving advantages are sometimes offset by the associated design modications
and benching requirements.
Today, model shops commonly use cardboard composites for rapidly
making larger touch/feel automotive prototypes like interior door panels,
instrument panels, and structural body parts. The labor-intensive process has
been accelerated by utilizing CAD part data (wire frame or surface) to generate
tool paths for driving two-axis NC knife cutters. Automated cutter machines
are used to make templates from cardboard sheets which are assembled to
form a 3D egg crate support and original part model. The part-model template is serrated to conform to the desired part prole when attached to the
egg crate support. The cardboard model is soaked in polyurethane and
removed from the support after curing. These models look very lifelike
and can be made in half the time required for traditional clay or cardboard
models. When needed, these models are used to make silicone molds for casting more durable polyurethane touch/feel prototypes.
B. Functional Parts
Fit/function prototypes are commonly used to verify t in assembly with
other parts and withstand some functional tests in the later stages of the
vehicle-design cycle. They are usually made from a material with properties
similar to the specied production material and must be dimensionally more
accurate than concept models. Traditionally, they are made in small numbers
198
Anderson
IV.
199
Figure 3 Low-volume fabrication tool grid. Soft/bridge tools (epoxy, kirsite, Al)
and hard tools (steel, Ni, ceramic) are either machined directly or made indirectly
from patterns by either a cast, thermal spray, or deposition operation.
200
Anderson
num core boxes (2 weeks versus 10 weeks). A single polyurethane core box
can make over 1000 sand cores.
B.
Plaster/Sand Molding
The die-casting process is one of the most economical ways to make a variety
of metal automotive components. Unfortunately, tool steel molds (with associated high fabrication cost and long lead times) are typically used to resist the
erosive hot metal ow and thermal fatigue encountered when making die-cast
parts. The associated time and money make prototype development for die
casting difcult. Thus, for eventual die-cast parts, an alternate prototypedevelopment strategy must be employed to effectively test and validate component designs within required budget and schedule constraints. The conventional prototype processes commonly employed for die casting are gravity
casting, machining from die castings with similar shape (when possible), and
machining from wrought or sheet stock. Among these three approaches, gravity casting has the greatest potential for die-cast development using RP&M
technology. Unlike machining, gravity casting is economical for low-volume
quantities and short lead times. After heat treatment, gravity-cast metal prototypes have properties (like surface nish, yield strength, and ductility) that
approximate die-cast parts. RP&M models have been successfully used as
patterns for making low-cost sand and plaster molds. These molds require
simple core designs and parting planes to remove the patterns from the mold
before casting. High-quality plaster and precision sand molds have been fabricated to gravity-cast, thin-walled, die-cast aluminum prototype parts (control bodies and throttle bodies) at a 60% cost savings over traditionally machined prototypes.
C.
Investment-Cast Molding
For more complex part designs, investment-cast ceramic molds (made by either Flask Casting or through the QuickCast process) have been used to as
a secondary operation to gravity cast prototype metal parts. The investmentcasting process (best suited for ne detail and close dimensional tolerances)
can be readily modied to make its use ideal for the rapid prototype development of complex die-cast parts. The modied process (a derivative of the
conventional lost-wax investment-casting process) incorporates the use of
cross-linked photopolymer QuickCast models as expendable patterns.
These patterns are burned off during the ceramic-shell mold-making process,
201
instead of being melted away as with traditional wax patterns. No parting lines
or cores are needed because the QuickCast patterns are burned out of the
mold cavity. As a result, the soft-metal (aluminum) tooling required to make
traditional wax patterns is simply not needed. These advantages simplify the
prototype-development process for complex parts or where conventional mold
designs (with many cores and parting lines) become cost prohibitive. For example, aluminum (SAE 356) prototype reactor blades for automotive torque
converters were made within 3 weeks instead of the required 30 weeks for
conventionally machined prototype reactor blades. The parts satised the specied tolerance of 0.02 in.
Furthermore, a tolerance of 0.002 in. (approximately 0.05 mm) have
been obtained for small parts by investment casting. Parts can range in size
from 1 to 36 in. The larger the part, the greater the tolerance required by the
process. In general, smoothness and accuracy becomes increasingly difcult
to obtain as the size of the casting increases.
D. Spin Casting
Spin casting can be used to make plastic, wax, or soft-metal prototypes in
sizes smaller than 9 in. The spin-cast process consists of pouring molten metal
or liquid thermoset plastic resin into the center of a spinning (2001000 rpm)
vulcanized silicone mold. This rotation forces the material outward under centrifugal force, resulting in pressures of 1015 psi which distributes the material
throughout the mold cavity and expels any trapped air before solidication
occurs. Metal parts can be spin cast at 50 casting cycles/h. Plastic (thermoset)
parts can be cast at 1015 cycles/h. The number of parts that can be made
per cycle can range from 1 to 10 parts, depending on part size and mold size.
Surface nishes of 90 m root mean squared (rms) are possible, and casting
tolerances of 0.0050.008 in./in. can be maintained from part to part. The
two preferred silicone materials used for tooling in spin casting are roomtemperature-cured RTV and heat-cured vulcanized rubber. The RTV molds
can withstand temperatures as high as 600F, whereas vulcanized rubber
molds can withstand temperatures as high as 1000F. Unfortunately, vulcanized rubber molds are formed under pressures as high as 4000 psi and at temperatures as high as 400F. Thus, RP&M parts to be used as patterns must
be able to withstand this temperature. Common practice is to use pewter or
high-temperature plastic parts produced in RTV molds as patterns or submasters to create multicavity heat-cured vulcanized rubber molds. Any metal that
melts below 900F can be readily spin cast in vulcanized silicone molds.
202
V.
Anderson
Making tools for both prototype part development and production component
manufacture represent one of the longest and most costly phases in the automotive product-development process. The sequential approach to production
tool fabrication by conventional machining is characterized by long lead times
and high cost. As a result, current practice is to start tool fabrication long
before product design is complete. Unfortunately, late design and engineering
tool changes commonly occur, making tool-fabrication lead time unpredictable. For example, front and tail light reector molds may undergo as many
as 16 tool-design alterations before completion. To accommodate these
changes, the tool material must not only be relatively soft to readily remove
material but also must be weldable to add material when needed for design
changes. Conversely, the tool material must be hard enough to resist wear and
forming loads. Traditional tools (molds, dies, and xtures) are machined from
wrought tool steel billets. Tool steels like SAE 4340, H-13, and P-20 are most
often used as die materials in production because of the unique properties
obtained through alloying and heat treatment. As a result, traditional toolmaterial selection is usually a compromise of properties (machinability and
weldability versus wear and strength) affecting performance.
This machinability compromise can be minimized using RT&M fabrication methods. RT&M can have a signicant inuence on reducing productdevelopment cycle time and cost. Figure 4 shows potential lead-time
reductions of various RT technology categories for low-, medium-, and highproduction-volume applications. Time-reduction estimates were based on the
Figure 4 Rapid tool development. Rapid tool fabrication methods are categorized
for various part volumes and reect potential lead-time reductions by eliminating the
need for many traditional machining operations.
203
A. Direct/Indirect RP&M
Stereolithograpy (SL) molds have been used directly to injection mold thermoplastic prototypes in small quantities (20,21). RP&M technologies like 3D
printing, LOM, and SLS has been successfully modied to make RT directly
from metal or ceramic powders (2225). The 3D printing process involves
the selective coating of powder metal (PM) with a organic binder by ink-jet
spray. This process is repeated layer by layer, forming a 30-m particle-sized
PM preform directly from a CAD model without molds or forming dies. Recently, the process had been applied to making small injection molds with
conformal cooling lines for improved thermal management (26). For LOM,
the RT process involves the use of a modied Helisys 1015 or 2030 machine,
where a tape-cast powder metal or ceramic sheet is cut by the machines laser
beam to create each cross section. The cross sections are then stacked together.
The scrap material is removed and the laminated part is presintered to burn
off the wax binder. Final densication is obtained in a conventional sintering
furnace. The process has been used to make small alumina wear inserts for
composite injection molds and has shown potential for other RT applications.
For SLS, the RT process uses DTMs Sinterstation 2500 System, where
metal powders coated with a wax binder are joined together when heated with
the machines modulated laser beam. The resulting porous metal preforms are
inltrated with molten Cu in a batch furnace, creating a fully densied composite metal part with moderate shrinkage. PM steel powders have been used
in the SLS process to rapidly make small injection-mold prototype tools that
replicate the actual manufacturing process using production materials.
Also, RP&M technologies have been used indirectly as patterns for making molds. For example, low-temperature castable molds like epoxy, chemically bonded ceramics, and RTV silicone rubber can be readily used for making prototype parts (in low volumes). RT methods in this category are made
most effectively when RP&M models are used as patterns (27). For intermediate and high production volumes, RT methods can be effectively used to make
204
Anderson
B.
For intermediate and high production volumes, making a protective metal shell
with cast aluminum, epoxy, or cement backing has shown signicant promise
for reducing lead time by over 30% compared to traditional tool-fabrication
methods. This approach allows mold components to be quickly assembled in
a composite structure for improved performance, as shown in Fig. 5. During
assembly, the protective metal shell can be mounted to a prefabricated modular
steel frame with standard insert bushings and guides for ejection pins and
cooling lines. The molds components are held together by casting, in place,
a composite aluminum-lled epoxy (CAFE) or cement backing material which
supports compressive loads and transmits tensile loads to the frame. Unlike
traditional tools, these mold components can be fabricated concurrently and
assembled quickly to produce fully functional tooling.
For wear resistance, the molds surface can be readily cast from any
metal material using wax or QuickCast RP parts as patterns. For small tools,
shell patterns can be quickly made by any of the RP&M technologies for
investment casting the molds active wear surface to shape with minimal machining (31,32). However, dimensional limitations should be considered when
investment casting large complex steel shells. The investment-cast ceramic
205
Figure 5 Composite tool assembly features/components: (from top to bottom) wearresistant surface shell or coating (arc metal spray, NVD, etc.); castable backing or
support material (chemically bonded ceramic, epoxy, etc.); prefabricated inserts (ejection pins, cooling coils, etc.); prefabricated supporting frame (steel, aluminum, etc.).
mold is a thermal insulator and must withstand large thermal gradients generated without failure as the casting cools. These gradients increase in severity
as the metal thickness increases and as the ceramic shells cross section decreases. More heat must be dissipated in heavier metal sections to maintain
cooling rates similar to thinner sections. In addition, thinner cross sections in
the ceramic shell are less able to absorb and dissipate this heat. This uneven
cooling becomes more severe in larger castings, which result in nonuniform
shrinkage, dimensional errors, undesirable warpage, and distortion.
As an alternative to investment casting for making larger molds,
RP&M patterns can be used as mandrels for depositing wear-resistant Ni on
206
Anderson
a molecular level to form high-density-shell tools with great dimensional accuracy, ideal surface nish, and minimal distortion. This can be done by nickel
vapor deposition or electroforming. Other applicable shell-making technologies that can incorporate the use of RP for RT composite mold assembly developments are arc metal spray, bulk metal spray and EDM. Because machining
is minimal, tool life can be improved by selecting protective shell materials
with greater wear resistance and heat-checking resistance than conventionally
used tool steels. The most signicant of these shell-fabrication methods (for
making large tools) are described in relation to their applications to niche
market manufacturing.
C.
The arc metal spray process can be used to deposit low-melting alloys (below
800F) directly onto plastic RP&M patterns, forming a solid metal impression
of the pattern without machining. The process uses two spools of metal wire
as consumable electrodes which are fed through a spray gun. An electric voltage is applied across the electrodes, forming an arc which melts its tips. Molten
droplets are forced off the electrodes and cooled by a high-velocity airstream
which propels them toward the pattern substrate, forming a mechanically
bonded coating subsequent to impact and cooling. When properly sprayed,
the shell coating formed never reaches temperatures above 120F, making
RP&M patterns ideal for use as substrates. This process continues, layer upon
layer, forming a solid metal shell conforming to the prole of the pattern
shape. Arc metal spray tooling is generally produced by spraying a soft-metal
shell (kirksite, a zinc-based alloy) an inch thick or less and backing up with
CAFE or a ceramic (33). The pattern is then removed from the metal shell
after the mold is constructed. Unfortunately, the protective metal coating produced in this way is porous and is susceptible to aking and spalling during
service. As a result, spray metal molds are best suited for applications requiring low pressures and temperatures (vacuum forming, blow molding, RIM,
injection-molding polystyrene patterns, sand core boxes). Arc metal spray
molds can be made in a fraction of the time required for traditional cast kirksite
prototype molds when using RP&M models as patterns.
Unlike porous arc metal spray coating deposits, bulk metal spray processes are characterized by high deposition rates where porosity and residual
stresses are minimized (34). For example, simultaneous spray peening and
the Osprey process can atomize as much as 150 lbs/h and 200 lbs/min, respectively, of molten metal. In the former process, metal deposits are simultaneously shot peened to increase density and reduce residual stresses. Shot
207
208
Anderson
having nickel carbonyl gas ow over a heated mandrel. Heating the mandrel
substrate to temperatures between 110C and 190C results in a uniform layer
of nickel being deposited on the mandrel at rates between 0.002 and 0.030
in./h, respectively. The resulting Ni shell conforms to the shape of the mandrel, with excellent surface replication, uniform wall thickness, and low residual stresses. Uniform wall thickness is obtained by minimizing mandrel temperature variations. Mandrel materials with high thermal conductivity like
aluminum or copper are best for this application. Using the NVD process, a
complex Ni-shell mold 1.0 in. thick can be made within 34 h.
This process has been applied to making composite tooling at a fraction
of the time for making traditional cast and machined prototype kirksite molds
(35). For example, a traditional kirksite injection mold for a plastic automotive
instrument panel would cost about $1,000,000 and require over 30 weeks to
build. In contrast, NVD or electroformed composite mold assembly for the
same instrument panel would cost about $300,000 and require 14 weeks to
fabricate. In addition, a traditional kirksite mold can make no more than 50
injection-molded instrument panels before reworking, whereas the harder Nicomposite mold assembly can make over 10,000 parts before rework. This
higher-tool-life mold is as good as the more expensive P-20 steel mold traditionally used in production. The high tensile strength (198,700 psi), hardness
(R c 48), and melting temperature (2647F) of NVD shells make its use as a
mold applicable to many fabrication processes: compression molding of SMC,
high-pressure injection molding of thermoplastics, RIM of thermoset plastics,
and sheet steel hydroforming.
An NVD composite assembly injection mold was made for Fords SN95 (Mustang) instrument cluster lens. A traditionally machined P-20 steel to
produce an automotive lens tool cost about $120,000 and take 1822 weeks
to make. The NVD tool cost 30% less and had a lead-time reduction of 60%.
A 5/8 in.-thick NVD shell was formed over a mandrel CNC machined to the
desired shape. After the Ni shell was removed from the mandrel, Cu cooling
lines were mounted to its back and added to a premachined platen assembly
complete with ejection pins. Cost and timing could be further reduced by
replacing conventionally machined A1 mandrels with RP&M mandrels with
similar thermal conductivity. The mandrel material must be thermally conductive and withstand deposition temperatures of 350F. Proprietary epoxy/
graphite composite materials have been developed to meet these requirements
but have yet to be made directly by a RP&M process. Currently, epoxy/graphite composite mandrels are cast to shape at room temperature in silicone molds
(made from traditional RP&M patterns).
209
Computer-aided engineering simulation indicated temperature variations as small as 2F across the surface of the NVD tool. Traditionally machined steel molds vary 10F to 15F across the surface. This lower temperature variation minimizes part distortion caused by residual stresses, thus
improving overall part quality. Over 19,000 parts were successfully made on
a 500-ton injection-molding press. No die wear was noted. Mold temperatures
were held at 140F for a 48-s semiautomatic cycle time to simulate current
production cycles using conventional tools. The parts showed no warpage as
predicted by the CAE analysis. Lowering mold temperature to 120F during
the cycle reduced the cycle time to 30 s (a 30% reduction in cycle time), which
greatly lowers piece price. The low thermal mass of the Ni allows the tool to
be thermally cycled between 140F and 120F for the best conditions to make
a high-quality automotive lens. This kind of rapid thermal cycling is not possible using traditional steel molds.
Unfortunately, the NVD process is not readily available because of the
potential health hazards associated with using the very toxic metastable nickel
carbonyl gas. Even though Canadian companies like NTT and Mirotech have
made great progress developing and implementing safety features for the process, more work is needed before use of the NVD process becomes widespread
for mold fabrication.
Electroformed Ni-shell tools have been made to injection mold plastic
parts (36,37). Electroformed Ni shells have been produced by both ExpressTool,
Inc. and CEMCOM Corporation to replace NVD shells as the wear-resistant
face to their composite mold assemblies. Electroforming is commonly used
in industry as a metal-plating process and is readily available for making Ni
shells. Electroforming Ni shells involves appling an electrical voltage between
a Ni anode and a cathode (with the desired tool shape), suspended in a aqueous
Ni salt bath. The positive Ni cations in a plating bath are attracted to the
negatively charged cathode, plating the substrate cathode with a Ni-shell coating. The resulting high-density Ni shell conforms to the shape of the mandrel.
Mandrel materials for electroplating must be conductive, insoluble in the plating bath, and withstand deposition temperatures of 130F. Electrodes have
been sucessfully made for electroplating by coating nonconductive RP&M
models with a very thin conductive layer of silver or graphite (5 ms).
Unfortunately, deposition rates are low and wall-thickness variations are
great, making process implementation slow for composite mold fabrication.
Ford Research Laboratories is addressing these issues and a greater use of
electroforming for mold fabrication is likely in the near future. ExpressTool
has recently announced commercialization of an electroform-based composite
210
Anderson
E.
VI.
A case study was made on a small (6 5 1.5-in. envelope), injectionmolded, polyproplene plastic automotive part. This part was an interior cover
for the electric sideview mirrorleft and rightfor the 95/96/97 Ford Contour and Mercury Mystique (see Fig. 6). As a benchmark, lead times were
(A)
211
(B)
A. Prototype-Development Methodology
Figure 7 is a schematic diagram representing the traditional prototyping methodology for automotive product development. After the concept-design stage,
2D drawings are sent to model or pattern shops to make 3D touch/feel
prototypes. These models are used to verify design intent and as patterns for
soft-tool fabrication. If a design error is noted, the 2D drawing is modied
to show the design change and another touch/ feel prototype is made. This
designchange iteration continues until the design intent is veried. Once veried, t/function prototypes are made using soft prototype tools. Soft
tools are usually machined from near-net-shaped castings of low-melting
alloys like kirksite or aluminum. Prototypes made from soft tools are used
for test to verify engineering design specications. If the prototypes fail the
test, a nite-element analysis is sometimes considered before design changes
are made and the whole process is repeated again. This phase of the process
212
Anderson
Figure 7 Traditional prototyping methodology. Nominal lead time: 30 weeks, assuming one redesign.
is characterized by costly, idle time between process steps. When test results
meet specications, the 2D part drawings are tossed over another communication wall to a tool and die shop where hard tools are designed and
made. Hard tools are usually machined from wrought tool steel stock like
H-13, 4320, or P-20. These tools are used to make either fully functional
prototypes for preproduction trials or actual parts for production.
In contrast, Fig. 8 is a schematic diagram representing the RP&M methodology for accelerated prototyping. The attached diagrams show that using
RP 3D models both as patterns and as touch/feel design correction helps
reduce lead time from 30 weeks to 13 weeks (a 55% reduction) for developing
the injection-molded prototype design of the polyproplene sail part. In
other words, replacing traditionally made patterns with RP 3D models not
only reduces the lead time to make patterns but also allows for comprehensive
design evaluations (a time-consuming iterative process) to be made early in
the development stage for improved design quality in a fraction of the time
normally required for traditional prototyping. Once this iterative design process has been completed, t/function prototypes can be fabricated for product verication testing. With the conventional approach, the only opportunity
to make design corrections is further downstream after the prototype parts are
fabricated.
213
In addition, CAE integration into RP methodology occurs naturally because of the readily available computer software programs for automatic grid
mesh generation using solid CAD data. RP methodology incorporates the creation of a solid CAD model before 3D part models are made. Thus, the development of nite-element analysis (FEA) can occur simultaneously with 3D
model fabrication. Prototype tests results can be compared to FEA results to
verify the analysis. Once veried, the model can be evaluated by an optimization computer program to select the best part design instead of relying on the
conventional trial-and-error approach normally used. Future improvements in
CAD visualization packages and virtual-reality software will allow designs to
be more readily understood without the need for physical models. Also, growing condence in FEA stress analysis will replace the need for mechanical
and ow evaluation tests on physical components. Unfortunately, automatic
mesh generation is not always best. Depending on the features of the part
design, intervention by a skilled FEA modeler is often still required. Eventually, automatic mesh generation will more readily account for various design
features, making user intervention unnecessary. Until then, RP will continue
to be useful for accelerating the process of making functional prototype parts
for test evaluation studies.
214
B.
Anderson
Tool-Selection Table
Figure 9 shows tool-fabrication cost and time for various tool/material types.
In this table, the inuence of RP&M on traditional tool selection is summarized for injection molding the small polyproplene sail part described earlier. Based on projected tool life, fabrication cost, and lead time, the table
optimizes the number of parts to be made for various tool/material types. This
optimization table can only be applied to the specic part selected. Even
though the table shows a general trend for injection molding most plastic parts,
caution should be taken when using this table to extrapolate optiminal tool
selection for other parts with different design features, sizes, or materials.
Using the traditional approach, machining is the most economical way to
make less than 500 parts. Machined plastic prototype parts are not completely
representive of injection-molded parts. They can only be used for simple mechanical test evaluations. In practice, more elaborate prototyping evaluations
are delayed until injection molds are made further downstream in the productdevelopment process. As the desired part volume is increased to 3000, the use
of cast soft tools (kirksite or aluminum) for injection molding becomes the
optimal choice for plastic part fabrication. Cast soft tools cost less to fabricate
than hard production tools ($30,000 versus $60,000) even though lead times
are almost the same (1214 weeks versus 1618 weeks, respectively). For
Figure 9 Production volume versus tool type for injection molding a plastic interior
cover for the sideview mirror of 95/96/97 Ford Contour and Mercury Mystique. *Conservative estimates based on experience. **The value can be increased using metal
inserts in critical wear areas.
215
production volumes greater than 3000, the use of machined hard tools (tool
steels) for injection molding becomes the most economical choice.
Using the RP&M approach, cast silicone is the economical mold of
choice for making less than 30 parts. These molds can be used to cast polyurethane parts (a thermoset for which the durometer hardness can be adjusted to
match polypropylene) that cost less and take 60% less time to make. Like the
traditionally machined polypropylene prototype parts, cast polyurethane parts
can only be used for simple mechanical tests to verify design intent. To make
injection-molded polypropylene (prototype) parts that better represent production, the RP&M approach can also be used to greatly accelerate the fabrication
of composite injection molds. Fabrication times can be reduced from 14 weeks
to 6 weeks when compared to traditional machined tools. SL cross-linked
photopolymer mold insert (i.e., Direct AIM) can be used to injection mold
thermoplastics for low-volume applications. Care should be taken because SL
material is brittle and is prone to premature fracture. For production volumes
less than 1400 parts, composite tools with soft-shell surfaces can be used (up
to 300 parts for CAFE or epoxy molds and up to 1400 parts for arc metal
spray kirksite molds). For greater part volumes, composite tools with hardshell surfaces can be used. In the tool-selection table, a 5000 part tool life
was estimated for Ni-shell molds. Recent experience suggests that the tool
life for Ni shells may be as high as steel-shell molds (up to 250,000 parts),
making it the optiminal tool choice for high-volume injection molding.
VII. SUMMARY
To remain competive in an evolving global economy, the automotive industry
has pushed to institutionalize processes that provide speed to the marketplace.
Recent trends have involved the reduction of product-development cycle time
and manufacturing cost. In one way or another, these trends have involved
the utilization of computer-aided fabrication technologies for accelerating the
product and manufacturing development process. Efforts have been focused
on developments involving automated fabrication technologies like RP&M
for low-volume component manufacture that improve communication both
internally and within the supplier base. Benchmarking Fords injection-molded
polyproplene interior cover (for the electric sideview mirror of the 95/96/97
Ford Contour and Mercury Mystique) as part of a lead-time reduction
case study indicated a 55% reduction (from 30 weeks to 13 weeks) between
traditional and RP&M. Replacing traditionally made patterns with RP&Mgenerated patterns not only reduces lead time but also allows for compre-
216
Anderson
217
duced fabrication lead time and cost (a rapidly growing future trend). Using
RP&M technologies to accelerate the manufacture of tools will grow from its
current economical use involving patterns for forming both soft prototype
tools and bridge tools to the basis of making hard production tools directly
from CAD data.
A variety of fabrication methods can be integrated with RP&M for future
RT developments. For intermediate and high production volumes, RT advancements will be achieved by engineering a tools physical and mechanical
properties at the particle or molecular level. Technologies like 3D ink-jet printing, LOM, and SLS have been successfully modied to make RT directly
from metal or ceramic powders. Indirect use of RP&M patterns for making
PM molds have been applied by the Keltool process, PM casting, and PM
forging. Powder materials electrolytically coated with organic binders are being investigated by a number of advanced material-development companies
with some success. Companies like, Lone Peak Engineering, and Rapid Dynamics are using RT fabrication involving technologies. Making protective
metal shells with cast aluminum, epoxy, or cement backing has shown signicant promise for reducing lead time by over 30% compared to traditional tool
fabrication methods. Unlike traditional tools, each component of these composite mold assemblies can be fabricated concurrently and assembled
quickly to produce fully functional tooling. Because machining is minimal,
tool life can be improved by selecting protective shell materials with greater
wear resistance and heat-checking resistance than conventionally used tool
steels. Several fabrication processes integrate well with RP&M technology
and have shown promise for making protective metal shells: investment-cast
steels, (arc/bulk) metal spray, NVD, electroformed Ni, and EDM. The idea
of using optimal tool-selection tables for a parts RT technology application
suggests future development of a multidimensional matrix for optimal tool
selection (the process-driven engineered tool) for similar parts with respect
to common design features, sizes, and materials.
The real challenge is not whether RT can meet niche-car market demands but whether RP&M can be effectively integrated into the automotive
manufacturing-development process. Effective RP&M implementation will
require large investments and will make extensive inventories of currently
used capital equipment for material removal obsolete. This implementation
will require fundamental cultural change in our product-development system.
Our traditional system still relies on the use of 2D CAD part drawings and
CNC machining. Most small automotive parts are outsourced to suppliers on
a competitive basis. They have little incentive to invest, not to mention incorporate, these new technologies into their product-development process.
218
Anderson
Among U.S. OEM automotive manufactures, over 67% of the average program investment is tooling related. Large part stampings like panels (body
side, door, quarter, luggage), hoods, deck lids, fenders, and reinforcements
account for 30% of the production tools and 70% of the cost. A future trend is
to replace whole stamping assemblies with more complex plastic composites.
Although consolidating a sheet metal assembly into a single more complex
part takes longer, the total fabrication time and cost is far less than what is
required to form and join simpler designed stamped components together. In
addition, tooling requirements are much less severe than for traditionally
stamped parts, making RT methodology more applicable. A development
strategy is needed where U.S. OEM automotive manufacturers allocate resource efforts to RT for making large parts, leaving suppliers to focus their
energies on implementing RT methodologies for making relatively small parts.
Rapid prototyping and manufacturing developments rely on solid CAD
modeling and state-of-the-art computer-aided technologies. To date, only 8%
of our design work force use solid CAD modeling and an even smaller percentage know enough about RP&M to take full advantage of its capabilities. The
proliferation of RP service bureaus seem to have relieved the problem, but
more capability and capacity will be required before RP&M can effectively
supplement traditional NC machining for automotive manufacture. Crossfunctional team efforts will be needed for improved communication both internally and within the supplier base while reducing technology development
cost through multiple resource leveraging. In the 21st century, the automotive
market will continue to become more global. To survive, U.S. OEMs must
meet the demands of our competitive future by adapting a consumer-oriented
mind-set on a global scale and become more exible to change.
REFERENCES
1.
2.
3.
4.
5.
RJ Scheetz. Operating philosophy for low volume production. SME Report No.
MF88-164, 1988.
LE Zeider. Automatic process generation and the SURROUND problem: solutions and applications. Manuf Rev 4(1):5360, 1991.
LE Zeider, Y Hazony. Seamless design-to-manufacture (SDTM). J Manuf Syst
11(4):269284, 1992.
T Sakuta. Development of an NC machining system for stamping dies by offset
surface method. SME Report No. MS870657, 1987.
J Rohwedder. Urethane for sheet metal fabrication. SME Report No. MF750991,
1975.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
219
S Wise. Net shape nickel ceramic tooling from RP models. SME Rapid Prototyping and Manufacturing Conference Proceedings, 1996.
Y Hazony. Seamless design-to-manufacture of marine propulsers: A case study
for rapid response machining. J Manuf Syst 13(5):333345, 1994.
S Schoeld. Engineering research deployment teaching initiative: Reducing
design-to-manufacturing time. Proceedings of the 1995 NSF Design and Manufacturing Grantees Conference, 1995.
J Dauvergne. Methodology and new tools for reduction of lead-time in product
engineering example of HVAC development. SAE Internal Congress and Exposition, Detroit, MI, 1994; SAE Report No. 940885, SP-1035, 1994.
GR Glozer, JR Brevick. Laminate tooling for injection moulding. Proceedings of
the Institution of Mechanical Engineers (IMecE), Part B: Journal of Engineering
Manufacturing VQ07nb1 1993, p9-14 (0954-4054 PIBMEU), IMecE, 1993.
DF Walczyk DE Hardt. A new rapid tooling method for sheet metal forming
dies. Proceedings of the Fifth International Conference on Rapid Prototyping,
Dayton, OH, 1994, pp. 275289.
DF Walczyk, NY Dolar. Bonding methods for laminated tooling. Solid Freeform
Fabrication Symposium Proceedings, Austin, TX, 1997, pp. 211221.
RC Soar, A Arthur, PM Dickens. Processing and application of rapid prototyped
laminate production tooling. Proceedings of the 2nd National Conference of Developments in RP&T, Buckinghamshire, U.K., 1996, pp. 6576.
FZ Shaikh, R Novak, J Schim, B Stroll. Precision stratiform machining: 100day engine project. Prototyping Technology International 97, 1997, pp. 286
291.
T Nakagawa, M Kunieda, L Sheng-Dong. Laser cut sheet laminated forming
dies by diffusion bonding. Proceedings of the 25th International Machine Tool
Design and Research Conference, Birmingham, U.K., 1985, pp. 505510.
P Dickens, D Sikon, R Sketch. Laminated tooling for moulding polyurethance
parts. Proceedings of the SME Conference on Rapid Prototyping and Manufacturing, Dearborn, MI, 1996.
M Grifths. Rapid prototyping options shrink development cost. Mod Plast
70(9):2427, 1993.
S Odette. Complex assemblies from stereolithography and RTV tooling. SAE
International Congress and Exposition, Detroit, MI, 1992; SAE Report No.
920744, 1992.
BA Jenkins. Epoxy tooling: Tomorrows hopes are todays realities. SME Report
No. EM860101, 1986. [Paper No EM 860101]
PF Jacobs. Recent advances in rapid tooling from stereolithography. Proceeding
of the 2nd National Conference on Rapid Prototyping and Tooling Research.
Buckingshire, U.K.: Buckingshire College, 1996. Mechanical Engineering Publications.
S Rahmati, PM Dickens. Stereolithography injection moulding tooling. 1997.
Rapid Prototyping Journal v3n2 MCB Univer Press Ltd Bradford Engl p53-60
1365-2646 RPJORC.
220
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
Anderson
E Sachs. Three dimensional printing: Rapid tooling and prototypes directly from
a CAD model. Proceedings of the 1992 NSF Design and Manufacturing Systems,
1992.
CW Grifn, J Daufenbach, S McMillin. Desktop manufacturing: LOM vs pressing. Am Ceram Soc Bull 73(8):109113, 1994.
S Michaels, EM Sachs, MJ Cima. Metal parts generation by three dimensional
printing. Proceeding of the 4th International Conference on Rapid Prototyping
and Manufacturing, 1993.
DL Bourell, RH Crawford, HL Marcus, JJ Beaman, JW Barlow. Selective laser
sintering of metals. Manuf Sci Eng 68(2):519528, 1994.
E Sachs, S Allen, H Guo, J Banos, M Cima, J Serdy, D Brancazio. Progress on
tooling by 3D printing; conformal cooling, dimensional control, surface nish
and hardness. Solid Freeform Fabrication Symposium Proceedings, Austin, TX,
1997, pp. 115123.
K Gottschalk, V Cariapa, G Wick. Feasibility of stereolithography as an alternative to prototype patterns for high speed sand casting. AFS 99th Casting Congress, Kansas City, MO, 1995.
N Hopkinson, P Dickens. Thermal effects on accuracy in the 3D Keltool process. Solid Freeform Fabrication Symposium Proceedings, Austin, TX, 1997,
pp. 267274.
H Noguchi, T Nakagawa. Rapid tooling by powder casting transferred from RP
model: Manufacturing conditions pursuing zero shrinkage. Solid Freeform Fabrication Symposium Proceedings, Austin, TX, 1997, pp. 287294.
AT Anderson. Rapid tool fabrication by powder metal forging. SME Rapid Prototyping and Manufacturing Conference Proceedings, 1997; SME Report No.
MF97194.
DA Van Putte, LE Andre. A step-by-step evaluation of building an investment
cast plastic injection mold. SME Rapid Prototyping and Manufacturing Conference Proceedings, 1995.
R Dzugan, RN Yancey. Investment cast tooling for metal casting and plastic
injection applications using rapid prototyping. SME Rapid Prototyping and Manufacturing Conference Proceedings, 1997.
LE Wise, EL Grusoz, FB Prinz, PS Fussel, S Mahalingam, EP Patrick. A rapid
manufacturing system based on stereolithography and thermal spraying. Manuf
Rev 3(1):40, 1990.
R Gansert. Near-net shape manufacturing by plasma technology. Proceedings of
the 1995 NSF Design and Manufacturing Grantees Conference, pp 423424,
1995; SME Report No. B2382423.
A Mathews. Nickel vapor deposition tooling for the plastics industry. Proceedings of the Third International Conference on Advances in Polymer Processing,
1993.
JR Logsdon. Electroformed nickel tooling. SME Report No. TE880215, 1988.
P Jacobs. New frontiers in mold construction: high conductivity material and
conformal cooling channels, SMR CMnCM 99115, 1999.
10
Rapid Tooling in the Medical Device
Industry
Daniel L. Anderson
DePuy Orthopaedics
Warsaw, Indiana
The signicant problems we face cannot be solved at the same level of thinking
we were at when we created them.
Albert Einstein
I. INTRODUCTION
We have all heard the saying, problems are opportunities in disguise. Problems do, of course, present the opportunity to nd solutions and, according
to Einstein, require an entirely different level of thinking. Unfortunately, people often tend to look for solutions much more diligently when there is an
urgent problem to be solved rather than simply planning ahead. Many facets
of industry are currently faced with problems/opportunities in the forms
of cost constraints, stiff competition, and reorganization of entire market segments. And, of course, once a competitor nds a better way, the better way
soon becomes the new standard.
The health care industry is facing the same challenges: providing the
best patient care possible while facing cost constraints from several different
directions. Health care providers are forced to limit available funds and related
services and/or share the expenses with patients, employers, the government,
or private sources. Competing successfully in this global environment, where
nationality, surgical expectations, and government regulation may dictate
222
Anderson
224
Anderson
225
age or are now affordable for implementation. The medical industry, particularly orthopedics, has certain applications, which, although not necessarily
unique, are extremely well suited to these technologies and may foster their
integration and expansion. This chapter will discuss the ways that RP&M has
enhanced the product development and the product launch phases of projects
in the orthopedic eld.
DePuy utilizes a series of RP&M systems that include stereolithography
(SL) and fused deposition modeling (FDM).
226
Anderson
the health care upheavals of recent years. We have denitely had to change
our level of thinking. As pressure from the competitive market and healthcare-type reforms increase, we search for ways to design and produce products
that are better and more cost-effective. We must work smarter. We are not
simply competing for market share and position, but, rather, have at the very
heart of our existence the desire to offer the latest and most technically capable
devices to our customers.
227
nd yourself going through the majority of the process again, with related
costs.
So, if tooling is so costly to generate, why use investment casting? Well,
there are basically four areas that must be considered: quantity, design, material, and speed.
Quantity. If you have to produce a large number of parts, then investment casting is often a very cost-effective mass-production
method. The up-front cost of the wax pattern tooling is nicely
amortized. However, if only a small number of parts are needed,
it generally is not good business to dump a large amount of money
into tooling; your per-piece cost will probably be disproportionately high.
Design. Generally speaking, the more complex the design, the more
machine and/or assembly time will be required to produce the
product. Fabrication of the complex shapes required for orthopedic joint replacements would require many hours of surface
machining. Investment casting can often be a cost-effective
method to produce complex partseven for a relatively low
number of partsif, of course, the up-front cost of the tooling
can be offset relative to the cost of the alternative. Again, if the
product is complex, the tooling will usually be complexand expensive.
Material. Some materials are much more difcult to machine than others. For example, cobaltchrome polishes nicely and interacts well
with UHMWP bearings, but its material properties make is less
than pleasant to machine. Again, investment casting can provide
some relief, if you can design castings such that there is little nishing work required to produce the end product.
Speed. Simply put, sometimes you can live with the lead times required
to develop wax pattern tooling, and other times you cannot.
Most manufacturing situations require a combination of these factors to
reach a satisfactory production decision. For example, the vast majority of
DePuy products that are made via investment casting are cobalt chrome and
consist of geometric shapes that would require extensive material removal if
machined. But the quantities vary. Sometimes, we need a small number of
implants for a clinical study; or, in the case of a patient-specic or custom
implant, we may need just one.
228
IV.
Anderson
If tooling were substantially less expensive and faster, or if there were a way
to produce accurate patterns quickly and cost-effectively without tooling, what
would be the impact on industry? Or more specically, DePuy?
Low-quantity casting runs could be more readily utilized for custom
implants, regional products, and clinical studies.
Lower overall casting costs could increase prot margins, reduce the
cost of the end product to the customer, or both.
Functional rst-article castings could be obtained much faster for
debugging nishing operations and/or to speed up product launches.
A.
229
230
Anderson
that will not melt) and there were some problems with shells cracking during
the melt-out phase of the investment-casting process, resulting in the possibility of corruption in the castings.
V.
CASE STUDIES
There are basically two categories where we use QuickCast: product launches
and custom implants.
A.
The following case study of a hip implant system shows how rapid tooling
impacts the product-launch process. The entire system includes 28 stems with
various neck angles and sizes, consisting of both right and left designs. The
goal here was to produce several of the intermediate sizes in order to facilitate
a clinical launch of the stem design. A clinical launch refers to the implantation
of several stems by the designing surgeons, in order to get a real-life feel
for the product and perhaps suggest last-minute improvements. Consequently,
design changes evolving from the study were a very distinct possibility.
There are three basic reasons why we took the rapid tooling approach
in the clinical launch of this product:
1.
2.
3.
Reduced cost
Reduced lead time for tooling
Changes from information obtained during the clinical period would
probably result in modications to the product
Cost is everything. Well, okay, so it is not everything. But nearly every consideration can be traced to cost concerns: cost of the product (of course), cost
of market share and sales lost because of delayed launch, and cost of selling
a product that did not have that one last opportunity for improvements. Each
of these cost issues are related to the three reasons listed.
Let us take a look at a hypothetical example of how RP&M impacts the
amount of time needed to launch a single hip implant. In this example, the
lead time for the conventional machined wax pattern tool is 12 weeks. After
the completion of the tool, it will take another week to get castings, and another
2 weeks for postprocessing. Furthermore, the time necessary to rework the
tool for minor design changes is typically about 4 weeks.
Given this information (and simple arithmetic), it can be seen that it
will be 13 weeks before the rst castings are completed and, consequently,
231
that long before manufacturing even gets to have a look at them. The customer
gets access to the product after 15 weeks, and if the customer or the manufacturing team wants to make changes, the time needed for tool rework kicks in.
At that point, the decision must be made whether to continue production with
the existing castings and phase in the changes, or completely halt production
until the new castings are available. If the decision is made to wait on the
new castings, the nished product would not be available for 19 weeks or
about 4.5 months! The key point is that the interval needed to generate and/
or rework a tool will almost always add large chunks of time to delay the
release of a product. Although this example is hypothetical, it is not too far
off from the actual case study.
In an ideal situation, tooling would be generated in a few days and castings would be available in 23 weeks, allowing Manufacturing an earlier opportunity to begin working with them. Also, the modus operandi would be
inexpensive enough to allow design improvements without turning everybody gray. This is exactly what RP&M offers (but with no guarantees against
going gray). Another added perk is the condence of seeing the rst wave of
the new design before committing to spending signicant funds for production
tooling. (This reduced commitment is often a motivation for the expedient
approval of a design.)
B. The Situation
The designing surgeons believed that this hip replacement system was an excellent product and they wanted to have access to the implants as soon as
possible. We believed that this hip replacement system was an excellent product and we wanted the surgeons to have access to the implants as soon as
possible. The designing surgeons like to have some exibility to make design
changes, based on the knowledge that they gain during the rst several implant
surgeries. We also like the surgeons to have some exibility to make design
changes, based on the knowledge that they gain during the rst several implant
surgeries. However, we want to be able to launch a product in a period of
time that matches business objectives.
The manufacturing group also likes to have some exibility. As seen
in Fig. 10, this implant required considerable postprocessing. Based on the
knowledge that Manufacturing personnel gain from the rst several castings,
changes to the design are possible to improve manufacturability and reduce
the nal cost of the product.
Thus, the issues basically boiled down to two: we need implants fast
and we need the exibility to make improvements without having to throw
232
Anderson
Fig. 10
expensive and time-consuming hard tooling into the recycle bin. By the time
the rst several castings had been implanted, we had good feedback on possible design tweaks from Manufacturing and the surgeons. It is clear that
rapid tooling increases your ability to take risks, as the tooling is neither prohibitively costly nor likely to seriously delay a product launch.
VI.
The development process does not drastically change simply because rapid
tooling is used rather than conventional tooling methods. The implant still
must go through its design phase, prototypes must be generated, and so forth.
Where the greatest impact takes place is in the initial launch of the product
and the decisions of when (or in some cases, if) it is appropriate to replace
the rapid tooling with production tooling.
The actual process to create the tooling is really rather straightforward,
but not as seamless as one may like. For instance, it would be great if all you
had to do was design the implant, produce an SL model of it, and pour a mold
around it. However, you nd out quickly that you must start dealing with such
issues as pattern and metal shrinkage, sacricial material, gates, and so forth.
Even these issues are not terribly burdensome if the proper approach is taken.
Fig. 11 Solid CAD model of casting design for the hip stern.
233
234
Anderson
Fig. 13
A.
The following case study for a custom femoral knee implant shows how direct
pattern generation can impact the production process. The major goal is to
reduce both cost and time to a level where a custom knee implant can be
generated cost-effectively and without negatively impacting other projects.
If the only option were to create traditional machined wax pattern tooling, then the cost of the project would be prohibitive, as the level of complexity
of the impact design would necessitate a four-piece wax pattern tool. One
Fig. 14
235
236
Fig. 17
Anderson
237
238
Anderson
239
Fig. 20 Solid CAD model of casting design for the knee component.
Finish the casting to match the implant design [Fig. 21 (right) and
Fig. 22]
As you can see, wax pattern tooling, the costliest step, is completely
absent from this procedure. Regardless of the method of production, the rst
two steps will have to be completed. If the production method were to generate
alternative tooling or machine the wax patterns, it would still be necessary to
do the rst three steps.
A solid model of the implant was designed in CAD using the established
design criteria. A solid model of the implant casting was then created, adding
material for polishing and nishing. The casting pattern was then created in
CAD by scaling the model to compensate for the shrinkage of the cobalt
chrome and, nally, adding gates. The latter information was obtained from
the casting vendor. The next step was to generate QuickCast models of the
casting patterns via an SLA. The QuickCast models were sent to the casting
240
Anderson
Fig. 22
Fig. 23
241
vendor, and the castings were received about 2 weeks later. The castings were
then nished, with one to be used as the implant and one as the trial.
The surgery took place early in 1995; the patient has since displayed
the results of a very successful surgery. The implant, by design, did not require
any revision of the tibial component. The patient has the added benet of a
one-piece femoral implant rather than one consisting of multiple components
required to ll in the areas of bone loss (Figs. 23 and 24).
VII. CONCLUSIONS
These case studies are examples of how RP&M technologies make possible
the creation of products that can improve lives through enhanced preoperative
planning and custom implants. As the use of RP&M and especially rapid tooling becomes common in the medical industry, more and more applications of
this valuable technology will be discovered and utilized.
ACKNOWLEDGMENT
All photographs were provided courtesy of DePuy Orthopaedics, Inc.
11
The Role of Rapid Tooling in
Investment-Casting Applications
Thomas R. Richards
American Industrial Casting, Inc.
East Greenwich, Rhode Island
Hugo Lorrain
Howmet Aluminum
Laval, Quebec, Canada
Peter D. Hilton
Technology Strategies Group
Concord, Massachusetts
I. INTRODUCTION
Rapid prototyping (RP) and investment-casting technologies have been used
in conjunction with one another since the early 1990s for various purposes.
One use of RP in support of investment casting has been to form the pattern.
Early work was with wax patterns produced by selective laser sintering
(DTM Corporation) or ink-jet printing (Sanders Prototype, Inc.). An alternative, involving plastic patterns produced by stereolithography (3D Systems),
was attempted and subsequently rened. Early patterns caused cracking in the
ceramic shell when they were burned out. Producing patterns with a continuous surface and a honeycomb interior (3D Systems QuickCast) solved this
problem.
Today, RP technologies are used in conjunction with investment casting
for at least two purposes: (a) to produce patterns for use in manufacturing
244
Richards et al.
prototype parts or very small sample sizes and (b) to produce patterns for
tooling (molds) which can then be investment cast. The tools are then used
to produce investment-casting wax patterns or for other mold-related processes
such as injection molding. The uses of RP in conjunction with investment
casting are relatively mature and in commercial application at numerous casting rms.
This chapter contains descriptions of the use of RP in investment casting
at two rms. The rst, American Industrial Casting, Inc., is a relatively small
and very innovative company. It is an example of a lead user among smaller
U.S. rms and demonstrates cost-effective implementation with relatively
modest capital investment. American Industrial Casting provides investment
castings to several market sectors. Its parts are typically small but detailed,
with tight tolerance requirements. The second rm is Cercast, a division of
Howmet, which is one of the largest producers of investment-cast aerospace
components. Cercast uses RP to form patterns for prototype investment casting
of large, complex aerospace parts.
Tom Richards is the technology leader at American Industrial Casting, Inc. Hugo Lorrain is responsible for prototype investment casting at
Cercast. They each describe some results achieved by their respective companies.
II.
American Industrial Casting, Inc. (AIC) is a manufacturer of precision investment-cast parts. It produces castings for the aerospacedefense, electronics
communications, mechanical components, medical, and subminiature parts industries.
These parts are characterized by the relatively small size (typical part
dimensions usually fall within a 7-in. cube in solid molds, although parts to
24 in. are produced in shell molds). Their requirements include ne-feature
denition of the order of 0.003 in. radii, walls as thin as 0.011 in. with high
aspect ratios, and tolerances of 0.003 in./in. up to 0.5 in. and 0.005 in./
in. above 0.5 in.
American Industrial Casting, Inc. focuses on producing production
quantities of nished parts in nonferrous or ferrous alloys from hard tooling
produced in 1012 weeks. These parts are often intricate in detail and thin
walled. AIC assists customers in their design for manufacture and assembly
245
246
Richards et al.
than several prototype parts rapidly and cost-effectively was to use the RP
pattern (appropriately sized to account for shrinkages) as a pattern for casting
a berylliumcopper alloy master. This metal master was then used to produce
temporary molds of vulcanized rubber, room-temperature vulcanized polymer
(RTV), or epoxy. These molds were, in turn, used to create wax investmentcasting patterns for producing parts. Unfortunately, these transfer-molding
methods resulted in problems. In the order mentioned, dimensional variations
ranged from 0.060 in. to 0.015 in. to 0.005 in., geometric distortions
from severe to moderate to marginally acceptable, mold-building times from
hours to weeks to months, and costs ranged from $50 to $500 to $2500 or
more.
American Industrial Casting, Inc. set out to develop a different method:
one in which RP castings are produced for the injection mold components
from RP patterns. The approach is straightforward. The designer, starting
with his 3D computer-aided design (CAD) nal part design, is coached in
the creation of shells around his part. Parting planes are installed in such
a way that the shells can be removed from around the injected wax or plastic
pattern without being damaged. These individual tooling components are
built as solid objects, molded by AICs solid-mold process and cast in a
berylliumcopper alloy. The process takes only 2 days. The alloy is very
uid and duplicates every feature of the pattern down to the nest detail
and nish. The resultant metal mold components are assembled using conventional mold nishing techniques. Waxes are then injection molded. The
waxes are then assembled, either into solid molds for nonferrous castings
or into shell molds for ferrous castings. Solid molds are produced by pouring a slip of refractory investment material around the wax patterns in a
vacuum environment and allowing the slip to solidify within several minutes
into a solid mold, which is dried and red overnight and held at a suitable
temperature for pouring the next day. Shell molds are formed by dipping
the assembled waxes into a slip of refractory investment material and hanging up to dry under controlled conditions, successively adding layers of
ceramic, over a several-week period, to complete the shell mold, which is
red prior to casting. The benets of this approach are lower cost and more
rapid development of injection molds capable of producing reasonable
runs of hundreds of functional metal parts. The molds are actually capable
of producing hundreds of thousands of parts, limited only by the high cost
of hand injection molding. AICs ultimate goal is to use their process to
make and install die cavity insets into rapidly produced and economical
machine injection molds that are suitable for the production of hundreds
of thousands of waxes for producing precision investment-cast parts. In short,
247
AIC is using investment casting to make wax pattern tooling for investment
casting!
These benets result from the direct transfer of 3D CAD geometry to
a physical geometry, followed by the use of standard investment-casting technology to produce metal alloy cast parts as-designed. The use of RP&M patterns plus investment casting of the mold components substitutes for the more
traditional computer numerically controlled (CNC) machining of the molds,
which typically requires 10 weeks.
Given these concepts, we need to quantify the capabilities and limitations of the RP&M direct tooling process as applied by AIC. The Sanders
Model Maker System denes many of these capabilities and constraints. The
current model MM6B Model Maker Pros are able to hold in-plane tolerances
of 0.001 in. The Z-direction resolution is set by the layer height. Layer
heights can be selected between 0.005 in. (coarse resolution) and 0.0005 in.
(ne resolution) with corresponding impacts on build time. Build times involve 28 s per layer of xed time, plus build rates that vary from 0.02 to 0.40
in.3 /h, for 0.00050.005-in. layering, respectively. As an example, the cavity
for an intricate part about 1.5 1.5 0.75 in. might be contained within
mold halves each measuring 2 2 0.5 in. overall. If 0.002-in. slicing were
selected, a build rate of 0.18 cubic in.3 /h could be expected, plus a xed time
of about 28 s per layer for milling. Thus, the two mold halves could be built
in about 24 h. Add 48 h for the solid mold process and an injection mold can
be ready for assembly and nishing in only 3 days! Although comparatively
slow, the process runs unattended overnight and builds patterns that are accurate, resolute, and smooth (80100 RMS) on all surfaces. Patterns up to 6 in.
can be built on the Sanders MM6B. Sanders Prototyping Inc. is continuing
development of RP systems based on its technology as well as technology
renements, so we expect that, when you read this material, their machines
capabilities will have been improved from the numbers given here. AIC also
employs service bureaus using other RP processes for economical building of
larger patterns.
An example part produced by AIC using the Prototype the Tool process is shown in Fig. 1. The part shown resulted from a CAD model of a
diode box which had just been put into production at AIC using conventionally
CNC machined tooling. It was chosen as a fair challenge for a rst RP&M
tooling demonstration project.
Virtual Concepts Design at virtcon.com was engaged to produce both
a 3D CAD model of the part and of the RP&M tool components for the diode
box working from the customers 2D print. Figure 2 shows the diode-box
injection-mold components placed alongside one another for the building of
248
Richards et al.
Fig. 1 An example part produced by AIC using the Prototype the Tool process.
Fig. 2 Diode-box injection-mold components placed alongside one another for the
building of the patterns.
249
Fig. 3 Floppy disk containing the data for the mold components (left), the nal cast
mold components (center), two wax patterns produced from the tool (right center), and
a nal investment-cast part in aluminum A356 alloy (far right).
the patterns. Figure 3 shows the oppy disk containing the data for the mold
components (left), the nal cast-mold components (center), two wax patterns
produced from the tool (right center), and a nal investment-cast part in aluminum A356 alloy (far right).
Once the les were made available, the entire process producing the
RP&M tool required only 1 week. The rst 12 castings required another 2
days plus an additional day for heat treatment. Consequently, only 10 days
after receipt of part les, 12 functional, heat-treated, metal castings were
available for the customer. Only a few years ago, this would not have been
possible.
250
Richards et al.
replaced (a) multipiece sheet-metal assemblies, (b) hogouts and forgings, and
(c) composite structures, for cost savings and improved damage tolerance.
Although the benets of structural airframe castings are being realized in most
new xed and rotating wing programs, the number of new sample and tooling
programs have been limited due in part to lengthy lead times. Rapid prototyping has demonstrated a unique capability to provide certication hardware
while production tooling is developed and matured in a parallel effort.
According to airframe designers, a door substructure (such as the one
in Fig. 4) can be designed as a precision casting much faster than an equivalent
multipiece sheet metal fabrication. Part count reduction, lack of fasteners,
absence of custom shim stock, and elimination of an extensive bill of materials
make a complex casting easier to design and procure than a traditional builtup structure. Unfortunately, the time savings associated with casting design
is usually offset by subsequent lengthy tooling and manufacturing times. For
251
programs with adequate lead times of several months, structural castings are
highly competitive and often specied. Rush retrot or short fuse development programs, however, often pass on advanced casting technology, except
where rapid prototypes can be obtained. In many programs, the ability to procure rapid prototypes is the deciding factor to design, test, and certify castings
into a production program.
There are no theoretical size limits to the RP process, and structures can
be cast with similar wall thicknesses, strength, and size scale as production
casting hardware. Castings excel at delivering components of high complexity,
incorporating many next assemblies into one single component. Reduction
of machining, joining, and tolerance stack-up from multicomponents provides
for unique structures.
Concurrent industry developments in recent years have yielded larger,
more accurate, and smoother RP patterns, in addition to reliable casting technology to transform these patterns into metal hardware. A description of the
complete investment-casting process can be found in numerous literature
sources. The RP process substitutes a pattern quickly produced using stereolithography for the heat-disposable wax/polymer pattern normally produced
from a production injection tool. In bypassing costly and time-consuming tooling, the foundry engineer can use this RP pattern to form a precise ceramic
mold, followed by pattern removal, mold curing, and subsequent casting of
metal into the mold cavity. Advanced metal alloys and/or rapid solidication
techniques can be employed to impart special characteristics to the casting.
Ensuing heat treatment, straightening, and nondestructive inspection (NDI)
techniques complete the process and yield a casting for nal machining, surface nish, and assembly.
IV.
252
B
Fig. 5 Final products made via SL.
Richards et al.
253
254
Richards et al.
A.
Project Requirements
In the pioneering days of rapid prototyping, the foundry engineer was pleased
to be able to demonstrate a successful transition from polymer model to metal
casting without catastrophic mold-split, inclusion, or distortion problems.
Yields were often less than 50%. Today, the technology has greatly matured,
due in part to new RP model resins and pattern build styles and also with
specialty designed gating systems, purpose designed shell mold processes, and
unique mold ring techniques. Modern-day prototyping yields components
with a 95% success rate, having predictable soundness, mechanical properties, and dimensional compliance. The mastering of pattern size variation and
dimensional process optimization was the last of several key parameters to
be accomplished. Requirements for modern-day aerospace prototypes are remarkably similar to certication of production hardware and may typically
include the following:
B.
Chemical certication
Casting microstructure evaluation
Mechanical property certication throughout the part
Radiographic compliance for soundness
Pressure tightness in leak tests
Surface nish validation and compliance
Full 3D prole tolerance compliance
The main objective for Bell Helicopter Textron was to reduce the time-tomarket and compress the procurement cycle for purchased metal structures.
255
C. Program Risk
Cercast has mastered the efcient transformation of a lightweight polymer RP
pattern into a high-strength aluminum alloy casting with few technical risks.
Aside from concurrently managing part design for producibility and determining capable tolerances for the component, one critical step remains. Gating
technology and ceramic shell mold design, based on years of empirical design
rules and experience, will improve the ability of the foundry to cast the component successfully with good soundness. Typical production techniques from
traditional hard tooling often require several cycles of trial-and-error gate optimization to produce a defect-free and economical casting. However, prototype
time constraints often require a usable component to be produced the rst
time around. A strategy of conservative gating and custom-designed shell
mold system (which encourages directional solidication) has enabled the
foundry to achieve a remarkable success rate, with few remakes being necessary on most designs. This is not to underscore those producibility discussions
between partners on items such as tolerances, wall thickness, weld rework allowance, xturing, and inspection aids that are all necessary for a successful program. Even with these requirements the xtures and manufacturing aids will not
be as complex and costly as those required for high-volume production runs.
256
Richards et al.
12
The Future of Rapid Manufacturing
Peter D. Hilton
Technology Strategies Group
Concord, Massachusetts
There continues to be strong driving forces in industry to compete more effectively by reducing time and cost while assuring high-quality products and
services. Some of these forces which will drive technology development and
implementation in the area of rapid manufacturing are as follows:
1.
2.
3.
258
Hilton
259
Traditional
injection
molding
RapidTool
injection
molding
Tooling cost
Tooling
development
time
Tooling life
$60,000
1618 weeks
250,000 parts
$0.24
$20,000
67 weeks
5,000 parts
$4.00
Table 2
Process
Traditional
tools
Disposable
tools
Moldless
forming
Steps
Design
part
Design
part
Design
part
Design
tools
Design
tools
Design
process
Make
tools
Make
tools
Make
parts
Make
parts
Make
parts
Store
tools
Install
tools
Make
tools
Make
parts
Make
parts
Make
parts
Note: With traditional tooling, a design change requires a tooling modication which is costly and time-consuming. With disposable tooling, the
tooling design change is made on the CAD system and the new tooling is made, as without a design change. The result is little cost or time impact
associated with design change. With moldless forming, each part can be distinct at no additional manufacturing cost.
261
thousand pounds per square inch. The result is that soft tooling can be used
for higher-volume runs and that disposable tooling produced by stereolithography can be used in mold frames to form hundreds to thousands of parts.
These are metal powder parts and they require sintering after molding but the
result is solid metal parts. This AlliedSignal technology enables the following
paradigm shift: Metal parts can be produced using plastic tools as differentiated from conventional wisdom by which plastic parts are produced on metal
tools. Of course, one can use this technology as a means to produce rapid
(metal) tools.
More broadly, the desire on the part of product-development teams to
have real prototypes (i.e., prototypes made from the production material by
the production process) will drive continuing improvement of rapid tooling
(or prototype tooling) technologies.
This desire is not frivolous; rather, it is based on the goal of easing the
transition from design to manufacturing by verifying early in the productdevelopment process that the parts can be produced by the anticipated manufacturing process. Further, this enables the development team to judge the
tolerance capabilities of the fabrication process as well as to identify aspects
of the design that may be difcult to produce. They can then make modications to the product design or the processing to achieve robust manufacturing
(the ability to produce parts within the required tolerances with a high degree
of certainty). Robust manufacturing avoids high initial reject rates as well as
early eld problems. This conceptual approach to reducing quality problems
is formalized through the use of statistics by setting allowable failure rates
and designing the combination of the part and manufacturing process to assure
that they are achieved. The terminology six sigma, which was rst popularized by Motorola, refers to using the above approach to assure that the manufacturing processes include six standard deviations within the part tolerance
band. The consequence of this method is that out-of-tolerance parts should
occur at a rate of four per million parts produced.
As described in this book, there are several processes under current use
as well as continuing development for making rapid tools that can be used to
create real prototypes. These include direct stereolithography mold inserts,
the use of SL or selective laser sintering (SLS) processes to form patterns for
investment casting of metal mold inserts, 3D printing of ceramic shells also
for investment casting of mold inserts, the various powder-based processes
(e.g., Keltool), and those which involve deposition of a hard metallic layer
over an rapid-prototyping (RP)-based pattern. Each of these processes is in
limited commercial use today and development work is continuing on all of
them. Although it is difcult to predict winners and losers among these tech-
262
Hilton
nologies, we can condently predict that rapid tooling will mature and that
its use will spread over the next decade.
Georges Salloum wrote about process simulation in Chapter 2 of this
book. Present analysis and simulation calculations have signicant limitations
that must be overcome in the future. Three-dimensional simulations of complex processes (e.g., injection molding or investment casting) may require
hours to days on high-end computers. Further, the simulations while predicting
trends and providing guidance are generally not sufciently accurate to predict
actual behavior. The inaccuracies are the result of both inaccurate input information on material behavior and approximations needed to reduce analysis
times.
We believe that the development of computer software tools to support
product development will continue and result in decreased need for paper or
prototypes. Such CAD/CAE/CAM tools will continue to become more accurate and efcient as the power of desktop computers continues to increase.
The result will be very fast responses for very complex calculations (e.g.,
simulation of the coupled uid ow and heat transfer during the lling of a
mold). The approach to product development will increasingly include CAD
design, CAE analysis of performance, simulation (and optimization) of the
manufacturing processes, and CAM, all using a single database and closely
coupled. A bit further in the future, computing systems will be fast enough
to enable real-time intelligent manufacturing process control (i.e., the process
parameters will be monitored and compared to the optimal values as determined by the earlier analysis). The process will then be continuously adjusted
to minimize the difference between actual conditions and optimal conditions.
Alternatively, the processes may be managed by neural networks that enable
learning and process improvement over time. Eventually, integrated computeraided design, simulation, and control will enable combined optimization of
product design and processing conditions, followed by actual processing at
these conditions. The results should include product performance improvement, product-manufacturing cost reduction, low (or zero) manufacturing reject rate, and high product quality.
One specic area in which computer-aided process analysis will support
process improvement is mold temperature control. One generally wants the
mold cavity active surface to maintain nearly uniform temperature, independent of the particular process (injection molding, investment casting, etc.) so
as to minimize part distortion and residual stress buildup during forming. Further, rapid transfer of heat from the part causes rapid part cooling and allows
shorter processing cycles, saving capital and variable costs. Computer-based
heat-transfer analyses can provide guidance on mold surface temperatures in
263
264
Hilton
costly than the machining processes they substitute. Net shape process utilization is limited by the cost and fabrication time for the associated tooling. Reductions in both these factors will occur as a result of a combination of the
technologies described in this book. The tooling will be further enhanced to
contribute to process optimization through such factors as conformal cooling.
The net shape (molding) processes themselves will also become more efcient
through the use of computer-aided tools for process optimization, including
process modeling (as discussed) and potentially neural-net or related techniques for continually learning and process ne-tuning.