Professional Documents
Culture Documents
The Rapid Tooling Testbed: A Distributed Design-For-Manufacturing System
The Rapid Tooling Testbed: A Distributed Design-For-Manufacturing System
requirements, then he/she can adjust the environment on which it is based. Section 4
design to facilitate manufacturing without is devoted to the experiments used to test
compromising functionality. the GT DFM method and the RTTB system.
The overall context for the RTTB is shown Finally, conclusions are drawn from our
in Figure 1. The three main stages in the experiences.
product realization process of relevance to
this project include functional design,
design-for-manufacture, and manufacturing
GT DFM method
(tooling and fabrication). Design involvement
is separated from manufacturing involvement, Problem statement and approach
but the software tools and information The purpose of DFM is to ensure that
formats for the design and manufacturing parts can be manufactured to meet their
organizations overlap. The key question of requirements in the most cost- and
the project relates to the timing of the transfer time-effective manner. To accomplish this,
from design to manufacturing and the scope it is often necessary to modify the part
of the DFM stage. designs to facilitate the selected
Overall, our approach to answer the key manufacturing process in a better way.
question was to experiment with different This is, especially, true when building
timings of design-to-manufacture transfer prototypes that are to be the representative of
and by scoping the DFM tasks differently. production parts. For example, a prototype
We separated the design and manufacturing gear train is to be fabricated in a different
activities in such a way that the designer material and with a different process than
transfers as much information as possible the production gear train. Further, assume
about the design, its requirements, and design that the prototype gear train undergoes a
freedom, and then allow the manufacturer to functional testing with a concern on its gear
perform DFM, process planning (PP), and tooth stress. Then, gear dimensions must
manufacturing. Design freedom refers to the be modified so that the prototype gear teeth
design’s attributes that the manufacturer can have the same stress characteristics as the
adjust, and the ranges of the adjustments, production gear teeth under similar operating
in order to perform DFM. conditions. The process of modifying
The digital interface between the prototype parts’ dimensions is called GT. For
designer and the manufacturer denotes the the most part, it is necessary to compensate
information package that was transferred to for differences in the material properties
the manufacturer (Fernandez et al., 2002). (between production and prototype
We experimented with STL files, CAD materials) when fabricating prototype parts
models, and design decision formulations in RP technologies. The situation is more
with integrated CAD and FEA models as complicated in injection molding prototype
the digital interface. parts, since there are two manufacturing
In the next section, we present our DFM processes to be considered: the process to
method called “geometric tailoring” (GT). fabricate the mold, and the molding process
In Section 3, the RTTB system is presented, itself. Both processes must be considered
as well as the distributed computing when performing the GT.
123
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
Given this introduction, the GT DFM specified by the designer or the manufacturer.
problem can be stated as: Designer specified information includes
Given a part design, functional requirements, design a parametric CAD model of the part,
freedom, and time and cost targets for a set of constraints and goals on functional, geometry,
prototype parts, compensate for differences in cost and time characteristics, analysis models
mechanical properties of prototype materials and
production materials by modifying the part geometry,
for these constraints and goals, target values of
designing a mold, and designing mold fabrication goals, and preferences for the goals. Although,
and injection molding processes. the designer has a preferred value for the
design variables, he/she specifies a feasible
Consider that a functional prototype is
range (instead of a point) for these variables.
needed to test a property of the production
The manufacturer provides process variables,
part. Using the concepts from similitude
RP and production material properties,
(e.g. Buckingham P theorem), the “property
process constraints and goals, and analysis
of interest”, X, for a prototype part and a
models. Note that process goals do not affect
production part can be formulated as a
any of the designer’s goals in the GT problem.
function of a set of part dimensions (Cho et al.,
Also, cost and time goals are affected only
1999). Part dimensions can be divided into
two categories based on their effect on X. by the manufacturing process. However,
Assume that dimensions d1, d2, . . ., dk have the designer specifies the targets for these
a significant effect on X, while dimensions goals to ensure that the part meets its design
d kþ1 ; d kþ2 ; . . ., dn have a minor effect. The specifications. The listing of these target
principle of the approach developed in this values in the GT problem serves the purpose
research is to change the model (prototype) of transferring these requirements to the
dimensions dm, j (for k þ 1 # j # n), such manufacturing group.
that dm, i (for 1 # i # k) match as closely To formulate a DFM problem, the
as possible the production dimensions designer fills in the MPGT template with
dp,i ð1 # i # kÞ based on the design and the information described earlier. This
process goals and constraints. These incomplete MPGT formulation is then sent to
dimensional changes are referred to as GT. the manufacturer, who fills in the remaining
Continuing the gear example, the face width information. With the completed
and diametric pitch dimensions would have formulation, the manufacturer is now able
a significant effect on the gear tooth stress, to solve the DFM problem, performing GT
while the diameter of the hole for the shaft of the part design. Hence, the MPGT serves
would have a negligible effect. as the digital interface between the designer
and the manufacturer.
GT problem formulation The system variables in the problem are
For RP-produced prototype parts, the the part dimensions that the designer has
problem formulations are called identified as having a minor effect on
material-process geometric tailoring functionality and which are modifiable by
(MPGT)/RP. These problems are the result of the manufacturer. These variables can be
combining a problem that captures functional modified within the range specified by
requirements and a problem that captures the designer to obtain better functional
manufacturing and material capability. properties. Deviation variables correspond to
Problem formulations are based on the the deviation of goal achievements. Each goal
compromise decision support problem (DSP) has two deviation variables (corresponding to
formulation (Chen, 2001; Sambu, 2001), under and over achievement). Bounds for
which is an extension of goal programming the system variables are obtained by the
formulations. Compromise DSP formulation physical limitations on geometry variables.
model decisions, in which, typically, multiple The deviation function is a linear physical
conflicting goals must be met as much as programming form of the Archimedean
possible, while satisfying a set of constraints formulation of deviation variables
(Mistree et al., 1993). A set of system variables (Hernandez et al., 2001).
may be adjusted in order to meet these goals. It should be noted that considerable
The MPGT problem formulation template is coupling usually occurs between the design
shown in Figure 2. and manufacturing goals, such as the cost,
In the MPGT formulations, the system surface finish, stress or displacement goals,
variables, goals, and constraints can be since they are typically the functions of both
124
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
design and manufacturing variables. For sub problems, nominally design and
example, geometry variables modify part size manufacture, that are simpler to solve, and to
and hence, affect the build time and part cost. decouple these sub problems to the extent
Surface finish can be a coupled goal under possible. Then, the manufacturing problem is
two circumstances. Firstly, varying geometry solved first for various values of design
variables affects the orientation of surfaces in variables. Design of experiments (DOE)
the part and hence affects surface finish, and methods are used to specify the design
secondly varying geometry variables varies the variable sampling strategy. After the
best part orientation and hence affects the approximation models of process capabilities
surface orientation and surface finish. are developed, the design sub problem is
Coupled goals complicate the solution process. solved using these approximation models.
MPGT for rapid tooling. For the case where When a rapid tooling problem is being solved,
the prototype parts are injection molded, two manufacturing sub problems are
additional considerations especially, the developed, one for RP-PP of the mold and one
variation of molded material properties must for the injection molding process. The
be included. The material and mechanical sequence of steps is shown schematically in
properties of SL molded parts can be different Figure 3, where only one manufacturing sub
from those of production parts (Dawson, problem (rapid tooling process planning
2001). For example, tensile modulus and (RT-PP)) is shown for simplicity.
strength of parts molded in SL are typically Decoupling of the sub problems is
lower than the parts molded in steel, while attempted by rearranging the goals and
flexural properties are enhanced. These constraints that are the functions of only
mechanical property differences cause design or manufacturing variables. Surface
prototype parts to have different behaviors finish is an example of a design goal that is a
than production parts in their intended in-use function of manufacturing variables only;
situations. The MPGT/RT problem finish goals are moved to the decomposed
compensates for these mechanical property manufacture sub problem.
differences. It combines three problems: one DOE and response surface methodology
for functional design, one for SL fabrication techniques (Myers and Montgomery, 1995)
process design of the molds, and one for the are utilized for two purposes in this work.
design of the molding process. Evaluation of goals and constraints often
require time consuming analysis, such as finite
Solution procedure element analysis. Rather than embed FEA in
The solution procedure is based on the robust an optimization loop, surrogate
concept exploration method (Chen et al., (approximate) models of the part’s functional
1997). The solution approach is to behaviors are computed. Additionally, SL-PP
decompose the MPGT problem into smaller can require substantial computing times due
125
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
to the need for the adaptive slicing of CAD product realization environment (Xiao et al.,
models of parts or molds. As a result, 2001). Web-DPR enabled users to interact
surrogate models of manufacturability for with product models, perform GT, and
different part dimension values are built to explore the effects of changes in project
further decouple the problem. requirements through web browsers.
The results obtained from the Modified Communications between agents in
RT-PP problem are used to solve the modified web-DPR occurred using the events that
MPGT/RT problem. As mentioned, the were broadcast through event channels.
modified RT-PP results are used to build However, instead of encapsulating message,
surrogate models of process capability for control, and information within an event,
various settings of part design parameters. events contained only message/control
The modified MPGT/RT problem is solved information, while application content was
for each solution obtained from the modified routed through a separate data flow. Message
RT-PP problem and a selection (based on information was encoded using XML.
the objective function value) is performed to Interoperability of distributed objects was
determine the best of the obtained solutions. accomplished using Java-RMI. The web-DPR
OptdesX with a simulated annealing (SAN) framework is shown in Figure 4. Note the
algorithm is used to solve the problem. The separation of message flows (through event
calculation of surrogate models, deviation channels) and data flows (through the
variables and objective function are data vault).
performed in C++ code integrated with The RTTB system was implemented on
OptdesX. the web-DPR framework. A set of general
and RTTB-specific agents were integrated
using two event channels, one for design
Web-based distributed product activities and one for manufacturing activities.
realization environment The manufacturing event channel, shown
in Figure 5, will be described further,
A distributed computing environment is since it is relevant to this presented paper.
essential for the implementation, testing, An incomplete MPGT problem formulation
and deployment of the RTTB. Research is the input into the manufacturing event
efforts were focused on two major aspects: channel, as described in the previous
one is the development of a suitable section. The event channel’s agents are
computing framework, and the other is presented later.
modelling the information that flows through The “process and material selection agent”
the framework to enable design and (not shown) was implemented by extending
fabrication. Three versions of distributed the selection decision support problem (DSP)
computing environments were developed, (Mistree et al., 1993). The essence of the
along with three methods of information selection DSP is to rate a set of alternatives
modelling. against a set of attributes, then rank-order
The third computing framework was the alternatives. In this work, the alternatives
called web-DPR, a web-based distributed are candidate materials and fabrication
126
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
processes, while the attributes are derived process capability will be developed. After PP
from product requirements. A utility-based is performed, this agent builds response
selection decision support problem (U-sDSP) surfaces and forwards them to the GT
formulation and solution method were formulation agent.
developed and tested as part of this project After a part or mold is designed, it is
(Fernández et al., 2001; Rosen and Gibson, necessary to design the fabrication process to
2002). be utilized to make the part or mold. Our
After selecting the appropriate materials approach is to develop experimental and
and processes, DFM activities are typically analytical models that relate process variables
performed. We developed GT methods to measures of part quality. Two agents were
for both part designs and mold designs. developed to support this work. The process
To support GT, we developed SL PP planning (PP) formulation agent takes the
capabilities. Also, limited injection molding material-process GT formulation and
process design was also accomplished. The formulates the SL PP problem. The
GT planning agent designs the set of compromise DSP was used as the problem
experiments from which surrogate models of formulation. Appropriate PP constraints and
127
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
goals are derived from the GT formulation. The final two agents (GT formulation and
This agent supports a high level of user GT) enable the formulation and solution of
interaction in order to perform problem the integrated GT problems. The GT
formulation. formulation agent enables the user to
The third agent, PP agent, solves the manually formulate the problem. The GT
formulated PP problem. Based on the agent is implemented as a wrapper around
experimental work, fabrication process design OptdesX.
methods were developed that enable the
selection of appropriate process variable
values to achieve to build goals of
accuracy, surface finish, and build time. Experiments and results
We developed a PP strategy that consists of
three stages: We conducted GT experiments with many
(1) orientation, parts, five of which are summarized in
(2) slicing, and Figure 6. GT was performed on all part
(3) parameterization. designs; GT for rapid tooling was performed
on three of the parts. Process selection, early
Trade-offs among build time, accuracy, and vs late, indicates the extent of design changes
finish are made in each stage so that only the that the manufacturer could perform. For GT
most appropriate process plans are passed to purposes, we assumed that parts would be
the next stage. Three generations of SL fabricated on a SLA-3500 machine in
process design software tools were developed CibaTool SL7510 resin. Molded parts would
(Lynn-Charney and Rosen, 2000; West et al.,
be produced using SL tools in polystyrene.
2001). The latter two versions incorporated
Experiment specifics (variables and goals)
a new adaptive slicing algorithm.
are shown in Table I, along with GT results. It
Given that the input to the RTTB will be a
should be noted that in the goal formulations,
part or product design, it must be possible to
target values represent ideal cases that are not
design a mold for the part(s) that can be
always expected to be achievable. Goals values
fabricated using RP techniques. The mold
are provided at ideal, desirable, acceptable,
design agent accomplishes this. As
tolerable, and unacceptable levels, according
implemented, this agent is a stand-alone
to the linear physical programming
program that is only loosely integrated into
formulation method (Hernandez et al., 2001).
the RTTB (hence, the dotted lines connecting
Constraints are automatically added to the
it to the event channel in Figure 5). We
problem formulation that prevent goal
developed a library of mold insert CAD
achievements at an unacceptable level. For
models (ProEngineer, SolidWorks) with
suitable mounting and ejection holes, gate, four of those parts, GT worked very well, as it
runner, and sprue features. After fabrication, was possible to improve the performance
the inserts fit into standard mold bases for characteristics of prototype parts relative to
injection molding parts. The types of rapid the production parts’ designs. In the fifth case,
tools (molds) that we have studied include enough design freedom was not available to
solid SL inserts, SL shelled inserts that are enable a significant change in the part
backfilled with epoxy or low-melting-point performance. Each problem will be described
metals, and epoxy inserts that are cast into briefly.
rapid prototyped dies. For the gear train problem, the ring gear
We developed a new method of automated was subjected to GT. Three design variables
mold design that was suitable for simple were included: face width (W ), diametric
two-piece molds, consisting of core and pitch (P), and number of teeth (N ).
cavity, as well as molds with many additional A maximum tooth stress goal was formulated.
moving sections (Chen and Rosen, 2002a). The target value for stress was adjusted to
Our approach utilizes planar parting compensate for the mechanical property
surfaces, which tend to result in better differences between injection molded ABS
molded parts than would be produced if and SL-7510 SLA resin. GT of the ring gear
curved parting surfaces were used with was successful. A MPGT/RT problem was
rapid tooling materials. These methods and also formulated and solved. Ring gears were
algorithms are reported by Chen and Rosen injection molded and tested successfully for
(2002b). functionality.
128
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
129
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
The light switch cover plate was less sizes of features. In the camera roller, many
successful. Two design variables were small slots are used to minimize weight and
included, the width (A), and thickness (T ), of facilitate injection molding, while providing
snap fits on the cover plate. Force, deflection, enough stiffness. The number of rows and
volume, time, cost, and surface finish goals columns in the array of slots (NR, NC) were
were included. Layer thickness (LT), hatch design variables. Additionally, other features
over cure (HOC), and fill over cure (FOC) could be added, removed or moved, but these
were the manufacturing variables. Despite cannot be described concisely by variables.
providing fairly wide design variable ranges, The diameter of the hole in the roller end (D),
the snap fits could not be adjusted enough to a slot width (Wi ), and rib thickness (t) were
completely compensate for the mechanical the other design variables. Given that film is
property differences between ABS and wound onto the roller, a torsional loading
SL-7510. condition was applied and an angular
The third problem, the simple robot arm, deflection goal (Z Rot) was formulated.
worked very well for both the part and the Weight, surface finish, tolerances, and cost
rapid tooling GT. Three design variables were were the additional design goals. Surface
modelled, the outside diameters, D and d, of finishes of mold pieces were the
the arm ends and the arm thickness, t. The manufacturing goals. After solving the
design problem is to have the arm deflect MPGT/RT problem, many goals were met;
minimally, subject to a bending force that those that were not met were still well within
simulates dynamic loading. Stress, deflection, the acceptable range.
weight, surface finish, tolerances, and cost are Generalizing from these experiments, some
the design goals. For the rapid tooling summary comments can be made.
problem, additional manufacturing goals are MPGT for RP. Most designs that were
added for the surface finish of the mold pieces tested showed significant improvement in
and for mold life. Molds were fabricated on a matching production-like performance
SLA-3500 machine and 50 robot arm parts characteristics. We could integrate the design
were molded in polystyrene. Several of these and manufacturing models effectively by
parts were tested and measured. Dimensions using the compromise DSP formulation. The
were accurate to within 0.5 percent. Tensile integrated problem formulation was
strength differed by 2.6 percent, while elastic decoupled effectively into a RP PP-problem
modulus differed by 8.5 percent. Tolerances and a GT problem. DOE and response
and surface finishes met or exceeded surface methods worked well in modelling the
expectations. Altogether, we believe that this coupling between the problems. However, we
example successfully demonstrated that the found that the response surfaces did not
GT method effectively compensates for always fit well into the design space, but the fit
material property differences between could be improved if the design space size was
prototyping and production material. reduced. For the simple robot arm
For the fourth problem, the stress and (experiment 3), a two-stage solution
deflection behavior of a robot arm filled with procedure was used. In the second stage, a
truss structure was investigated. Again, a smaller design space was formulated, based
bending load was applied to the arm. The on the results of the first stage. A significant
design variables were eight groups of truss improvement in the prototype performance
element diameters (D1 - D8). A maximum was achieved.
stress constraint value was computed by In one GT problem (experiment 2),
compensating for the flexural strength significant improvements in prototype
differences between a mild steel and SL-7510. performance could not be achieved. This is
The MPGT/RP problem was solved because the design variables could not be
successfully, minimizing part weight while modified without violating other design
meeting the deflection target. requirements. The conclusion here is that the
For the fifth problem, the rapid tooling GT success depends on providing the
problem will be presented. We simulated the manufacturer with sufficient design freedom
early transfer of the problem from the designer to enable some DFM.
to the manufacturer by allowing the MPGT for RT. The MPGT method for
manufacturer to modify the number and rapid tooling was applied to three parts,
arrangement of features in addition to the namely, one gear in the gear train, simple
130
A distributed design-for-manufacturing system Rapid Prototyping Journal
David W. Rosen et al. Volume 9 · Number 3 · 2003 · 122 –132
Chen, Y. (2001), “Computer-aided design for rapid tooling: Hernandez, G., Allen, J.K. and Mistree, F. (2001), “The
methods for mold design and design-for- compromise decision support problem: modeling the
manufacture”, PhD dissertation, Georgia Institute of deviation function as in physical programming”,
Technology. Engineering Optimization, Vol. 33 No. 4, pp. 445-71.
Chen, Y. and Rosen, D.W. (2002a), “A region based Lynn-Charney, C.M. and Rosen, D.W. (2000), “Accuracy
method to automated design of multi-piece molds models and their use in stereolithography process
with application to rapid tooling”, ASME Journal of planning”, Rapid Prototyping Journal, Vol. 6 No. 2,
Computing and Information Science in Engineering, pp. 77-86.
Vol. 2 No. 2, pp. 86-97. Mistree, F., Hughes, O.F. and Bras, B.A. (1993),
Chen, Y. and Rosen, D.W. (2002b), “A reverse glue “The compromise decision support problem
approach to automated construction of multi-piece and the adaptive linear programming algorithm”,
molds”, ASME Journal of Computing and in M.P.K. (Ed.), Structural Optimization: Status and
Information Science in Engineering, Promise, Washington, DC, pp. 247-89.
(in preparation). Myers, R.H. and Montgomery, D.C. (1995), Response
Cho, U., Wood, K.L. and Crawford, R.H. (1999), Surface Methodology: Process and Product
“Error measures for functional product testing”, Optimization using Designed Experiments, Wiley,
ASME Design for Manufacturing Conference, New York.
Paper #DETC99/DFM-8913, 12-16 September, Rosen, D.W. and Gibson, I. (2002), “Decision support and
Las Vegas. system selection for RP”, in Gibson, I. (Ed.), Book
Dawson, K. (2001), “The effects of rapid tooling on final chapter in Software Solutions for RP, PEP Ltd,
product properties”, PhD dissertation, Georgia London.
Institute of Technology, Atlanta. Sambu, S.P. (2001), “A design for manufacture method for
Fernandez, M., Rosen, D.W., Allen, J.K. and Mistree, F. rapid prototyping and rapid tooling”, Masters thesis,
(2002), “Digital interfaces: the key to effective Georgia Institute of Technology.
decision-making in distributed collaborative design West, A.P., Sambu, S. and Rosen, D.W. (2001), “A process
and manufacture”, ASME Computers and planning method for improving build performance in
Information in Engineering Conference, DETC2002/ stereolithography”, Computer-Aided Design, Vol. 33
CIE-34466, 29 September-2 October, Montreal. No. 1, pp. 65-80.
Fernández, M.G., Seepersad, C.C., Rosen, D.W., Allen, J.K. Xiao, A., Choi, H-J., Kulkarni, R., Allen, J.K., Rosen, D. and
and Mistree, F. (2001), “Utility-based decision Mistree, F. (2001), “A web-based distributed product
support for selection in engineering design”, ASME realization environment”, ASME Computers in
Design Automation Conference, DETC2001/DAC- Engineering Conference, DETC2001/CIE-21766,
21106, 9-12 September, Pittsburgh, PA. 9-12 September, Pittsburgh, PA.
132