CONCURRENT DESIGN OF PRODUCTS FOR MANUFACTURING SYSTEM

PERFORMANCE

P. Soundar and Han P. Bao

Department of Engineering Management

Old Dominion University, Norfolk VA.

Abstract working with the most up-to-date description of the

Manufacturing cost forms a very major portion design to evaluate the effects of particular design
of the life cycle cost of a product. Aho, performance of featuresl."[l] This definition is well complemented by
the manufacturing system directly afects the the research that is currently performed in this field.
manufacturing cost of product produced in the system. "'bere is a whole group of researchers who have worked
lhe design of products in the system aho Gects the on capturing the knowledge of these individual specialists
performance of the manufacturing system. However, a and transforming it into the computer so that the designer
survey of research shows that the research community has can use them in designing the product. The research
igmrtd consideration of manufacturing system proposed here draws on both the uses of the term CEST.
performance during the design stage, lhis paper It allows specialists to evaluate the effect of particular
identifies critical research questions in this area. In this design features, at the same time allowing for the use of
paper we also propose a system that would allow knowledge bases to perform some of their tasks.
manufacturing system pdormance measures to be CEST when fully developed and implemented
considered during the design stage of a product. We at30 would allow the designer to evaluate the product design
address issues in the integration of manufaauring models features from the perspective of specialists of departments
with product design somare. Ihe system takm advantage like quality control, assembly, manufacturing, service,
of currently existing technology and allows for etc. This allows for the simultaneous consideration of all
integration of new developments as they occur. the life cycle phases early in the product design stage.
Eventually CEST should allow for the simultaneous
design of manufacturing facilities for the product.
1. INTRODUCTION However, the design automation tools developed to date
test the design features only for their manufacturability
Traditional manufacturing organizations follow using particular process operations. However, any set of
an "over-the-wall" approach to product design and tools developed must satisfy a desired set of
manufacture. Organizations that practice concurrent characteristics:
Engineering (CE) or simultaneous engineering consider 1. a coordinated set of user interface principles over
all the life cycle aspects of the product in the early design application areas
stage. Seven capabilities must be in place for an 2. easy to use even by people with little knowledge of the
enterprise to successfully implement CE [l]. They are: system
1. Supporting Management Practices 3. ease of communication among, and addition, of work
2. Empowered Multi-function Teams domains.
3. Customer-Supplier Support Most often cited benefits achieved using CE
4. Quality Functions and Formal Methods include:
5 . Concurrent Engineering Support Tools dramatic reductions in engineering design changes
6. Enterprise-Wide Information System and during the early part of the design
Environment drastic reduction in product development cycle time
7. Rapid Technology Deployment reducedscrapandrework
reduced manufacturing costs
The remainder of the paper focuses on one of These benefits of CEST justify further research in this
these capabilities namely, concurrent engineering support
tools (CEST). "CEST are computer based tools that
allow specialists from different disciplines the freedom of lsome examples of design features are: weight, material
used, length, height, breadth, surface texture, etc.

233 CH3458-719410000-0233 $1 .OP1994 IEEE

subject area. A mjor boost to the practice and research in CE
There is a rather large void in the support tools techniques was provided by DARPA in 1987 [4]. At that
that are available to enterprises desiring to practice CE. time DARPA was concemed about short-comings in the
This paper proposes to research two questions that would process of defeuse-related product development in the
fill the void. The performance of manufacturing system US. DARPA initiated the DICE (DARPA Initiative in
can be improved using a set of philosophies that Concurrent Engineering) mission to "develop, validate
emphasize realigning activities within the shop-floor and disseminate technologies that encourage the use of
namely, JIT, Continuous Improvement, etc. These are CE in the design and manufacture of defense-related
often combined with manufacturing system design2 to products." The focus of the project was to develop
improve performance. However, integrating a new advanced computer based communication and data
product line with existing product lines adversely affects exchange technology that would make possible the
the manufacturing system performance [3]. This is creation of "virtual tiger teams". The DICE initiative
because system parameters have to be changed to was not concemed with automating the design and
accommodate the new product. To avoid this situation development process. This was very much left to private
the performance of manufacturing system should be initiative.
explicitly considered in the design stage. The DARPA initiative facilitated the co-location
The design of the product, among other things, of remotely located teams, but CE is much more than
affects the performance of the manufacturing system. CE teamwork. Computer-aided design, engineering and
research in this area has been focused on testing the manufacturing tools initially played a major role. But
manufacturability of a product at design stage. The recently these tools have been expanded and extended to
concurrent design of products and performance of include modules for quality, reliability, serviceability,
manufacturing systems required to manufacture the and disposability. There are tools that even deal with
product has been largely ignored. Hence one of the environmental impact on designs. Since, these particular
questions that should be researched is "How to detect aspects are beyond the scope of this proposal such tools
product design features, that Significantly affect the will not be reviewed here.
performance of its manufaduring systems early in the In a special report on the benefits of CE, IEEE
design stage?" This would allow the product designer to Spectrum (51 presents four case histories of companies
incorporate manufacturing system performance that have benefited from the techniques of CE applied
considerations early in the product design stage. using CEST. CEST has even been the topic of a book,
The answer to question I establishes a "Concurrent Engineering Automation, Tools and
relationship between manufacturing system performance Techniques" edited by Andrew Kusiak.[6] This book
and product design features. This would lead to contains individual contributions from academia around
identification of design features that have the most impact the country who have developed various design
on the performance of manufacturing system. The second automation tools for practicing CE. The range of the life
question that should be reseprched is "How to determine cycle phases covered by these articles include structural
the values of the individual features that will optimize design, parameter design, tolerance design, assembly and
the design utility value for the designer?" manufacturing, quality, reliability, serviceability,
maintainability and economics. However, to maintain
2. LITERATURE REVIEW focus, only the articles dealing with design and
manufacturing are dealt with here.
Concepts similar to CE have been practiced b Colton [7] presents an Intelligent Computer
the Japanese for years through the use of "tiger teams6 4 Aided Design (ICAD) tool. The tool allows the designer
approach. This enabled them to have drastically lower to interactively change his design based on manufacturing
times from product concept to product availability than constraints. The product is modeled based on IDEFlX
their American competitors. In the US the time modeling methodology. The knowledge in this model
compression that can be realized from the practice of CE structure is based on semantic nets, i.e., in this modeling
techniques was realized in mid 80's. methodology model entities are linked using "has" or
"has-either" relationship. In addition, the model uses a
coexistence relationship. This allows the capture of
2manufacturing system design addresses broad issues of relationship betwem two mechanical design features that
factory organization, producibility , buffer requirements, can be represented as separate entities but cannot exist
and overall performance, such as throughput, flexibility, separately. For example bolt and thread have a co-
work-in-process and machine utilization [2] existence relationship.
L t a i l S gathering all the stakeholder in c m place at the The designer designs products at four levels, the
start of a project thst addnsses all the participants' feature level, part level, sub-assembly level, and product
individual and ther tenm's global concerns [q

234
level. For example to design an engine, he first designs presented in this article ensures that the manufactured
the individual features of the piston like the diameter and product has design features that are cost effective to be
length, then he combines these features to design the manufactured. However, this article falls short of
piston, next he proceeds to the piston piston-ring sub- providing a link between the design features and
assembly, and finally he combines these individual sub- manufacturing system performance.
assemblies to design the engine. The system checks for Another active research area in concurrent
the manufacturabiiity of the product at each level of the engineering is design for assembly. There are two
design. The manufacturing rules and heuristic are stored different groups of researchers in this area. The first
in a knowledge base. The system was developed using group of researchers feel that product assembly must be
the C programming language on a Macintosh IIX made simpler so as to facilitate automated assembly.[9]
computer. The design system presented ensures that the The argument is that traditional assembly operations
designed parts do not have features that cannot be require human dexterity and if this operation is to be
machined or use fastening techniques that are not automated, it has to be simplified. So the main focus of
appropriate. However no attempt is made to link these remanhem has been on simplifying design to
manufacturing system performance with product design. automate assembly. The focus of these tesearchets has
Dong [8] presents a method for optimizing also been on design of robot grippm and the forces and
design features for manufacturing costs. There exists pressure required to perform assembly operations.
several empirical models that define relationships betwea The second group of researchers, the most
tolerance and production costs. Some models that exist prominent of whom are Boothroyd and Dewhurst [lo],
are exponential, polynomial and reverse polynomial. are focusing their attention on the developments of
Given a set of p independent tolerances that can be techniques to reduce the assembly operation time. Today
adjusted from a set of n tolerances, the optimization there are software available that evaluate product designs
problem can be formulated as for assembly efficiency and suggests methods to improve
them. Their basic attention is focused on removal of
fasteners and reduction in the number of assembly
operations by combining sub-assemblies into a single
part. The software analyzes the part for ease of assembly
P
min (w.r.t.bi) CC,(S,)
i=l
as well as ease of material handling. The researchers
have developed standardized forms to analyze the design,
and the system also identifies the problem parts and how
Subject to they can be improved. The disadvantage of this approach
6,+6,+-------+ti, = 6, is that although it might take less time to assemble
products, it does not necessarily mean that it is easy to
and produce them. Once again only the efficiency of a
(i = 1,2,,.,.n) operation is considered but the effect on the
manufacturing system performance is not considered.
The link between design and manufacturing
where 61,---,6 are tolerances that can be adjusted, 6~ is exists in traditional CAD/CAM systems. However, in
the sum of alp tolerances in the part and ci("i> is the traditional CAD/CAM computer-aided tools are first used
production cost associated with individual tolerance i. to design the product and then the data is transferred
The optimization problem is solved using numerical electronically to manufacture it. Some systems contain
optimization. modules that automatically test the manufacturability of
In this work the author presents accuracy graph the product as designed. Even if a product is
representation technique that provides an association manufacturable it might still call for unreasonably long
between tolerance and feature surface that it represents. material handling and p d g times. Either scenario
It is a well kuom fact that mechanical features in a leads to additional costs and wastes. To overcome this
design are related to production processes. The relation problem Izuchkwu [11,12] suggests a method to detect
developed by the author leads to a connection between such problems in the early design stages.
mechanical features and production costs for making the The author strongly advacates the use of
feature. artificial intelligence tools to solve such problems. He
The quantitative manufacturing knowledge (like hypothesizes that there is a direct co-relation between the
cost-tolerance relations of production pmcesses) is stored shop-floor material handling time and the products design
in the knowledge base. Thus the link between features complexity. He suggests the use of transformation
and tolerances provides a direct link between features matrix, which is used by CAD systems, to obtain the time
and the costs to produce them. This allows for evaluation the product spends in the shop-floor. This he contends
of candidate designs for production costs. The technique

235
can be used to change the transformstion matrix and thus
alter the product design to optimize the manufacturing 1 ’
time. This seems like a reasonable approach, however, it Ui(X) = -{[rplcjuj(xj)
+ l)] - 1)
must be emphasized that this system has not been k j=1
implemented yet. where Vi@) is the overall utility of set of attributes X for
General Motors uses an feature-based design each design alternative i, i=1,2, ... ., I; U.(x.)
J J
is the
strategy to integrate the information generated in the single attribute utility function for attribute J; x, is the
design and manufacture of their power-train components. .
performance level of attribute j; j = 1,2,. . . ,J attributes;
The system is based on a database that provides the kj is the single attribute scaling coostant, and K is the
design engineers with up-to-dnte information on the latest n o d i z i n g constant derived from
processes and technologies available to the company for
manufacture of power-train components. 1131 This I
information is used for optimizing manufacturing 1 + K = n(l+Kkj)
processes, determining product and process interaction, j=l
and assessing competitive product strategies. What this The authors present a method for determining the single
system lacks is the capability to integrate manufacturing attribute utility functions and scaling constants.
knowledge from their own shop-floor with product
design. 3. RESEARCH PLAN
Simultaneous evaluation of several design
attributes including manufacturing costs, weight, quality This section of the paper discusses the general
and mechanical design considerations is presented by approach, research methodology and the procedures that
Thurston and Carnahan [14]. The problem that is tackled if adopted will provide a solution to the research
is the determination of the combination of performance questions. The research can be performed in two separate
attribute levels that provides the greatest over all value or phases. In the first phase the design of the system that
worth to an individual designer. The evaluation system will enable manufacturing system performance to be
the authors present is in two parts. The first part consists considered in the design stage will be tackled. The second
of an expert system module which is used to reject design phase addresses the integration of multi-attribute
alternatives that do not satisfy manufacturing constraints. optimization into the design system.
The second module ~ssesses the multi-attribute utility Figure 1 shows the system design that would be
function of the user. Traditionally multi-attribute design used to answer research question I. The figure shows the
is evaluated using a Figure-Of-Merit (FOM) value software modules and the information flow between the
determined by the following equation modules. Research will determine what are the specific
data requirements and data formats to be used in the
n system. Some of the modules will require human
FOM= Z[wi(- pi )I knowledge to interpret the data and make decisions .
i=l Pi- Although, human involvement in making the critical
wi is the weight of one of the n attributes, Pi is the decisions in the system will be significant in the
attribute level of a particular design alternative and beginning this will be graddly reduced as the system
Pi,- is the maximum level of attribute i exhibited by gains maturity.
all designs under consideration. In the absence of a linear Next the function of each module and how it
relationship between a users desire and attribute a Non- interacts with the rest of the system is discussed. Please
Linear Merit (NLM)measure is used refer to figure 1 as you read this section.
n System Outline Features and tolerances from CAD
NLM = Z[wivi(Pi)] systems are used to identify the p d g times and
i=l resource requirements of the product. The routing and
vi(Pi) is the measure of performance of attribute i. The material handling requirements are determined for the
summative multi-attribute utility functions can be used product and these are linked with the other activities
only if being performed in the plant at that time. Static and
dynamic modehg is performed on this data, and if the
results are not satisfactory the relevant featurea identified
t w i = 1
and the design modified. If the results are satisfactory the
i=l
If this is not the case then a multiplicative model must be tooling and fixtures required are made and the product
used. manufactured.

236
 CAD System The CAD system will be one of the
commercially available systeb that can perfom 3D solid
modeling. The use of a standard CAD package will
enhance the portability of the system. The design
features and tolerances are the two important CAD system
values passed on to the process identification module.
Features are any surface that requires machining or any
other manufacturing process to make it. Examples of
features are, holes, chamfers, grooves, slots, etc.

manufacturin Process Identijkation Identification of machining

database resources from design features has been a major focus of
resource research, for researchers in Automated Process P l h g
Systems [17]. Several algorithms and methods that
vprocessing times identify manufacturing requirements from design features
have been developed. These can be incorporated into the
other jobs process identifications module. Initially a human expert
scheduled will be involved in this module however, as techniques
routing,
improve this stage can be fully automated. Some features
can be easily linked to the machining resources required
to produce them. For example holes need a drill. And
MODEL then based on the tolerance specified it can be decided if
there is additional processing required, like boring and
reaming. However, to produce a flat surface it is more
dynamic queue difficult to determine if it requires shaping machine,
planing machine or milling machine. This module will

1
machine changes, dynamic
utilization, machine utilization, be updated as developments in this area occur. The
demand satisfaction flowtime, WIP process identification module interacts with the
manufacturing database. The manufacturing database
contains information about the various resources in the
shop-floor and processing times of the parts.

Master Scheduler Most enterprises have a computerized

master scheduler. The management sets the demand and
due-date for the new product. The vital output of the
design master scheduler is information on other products to be
fmtms, produced in the manufacturing system at that time. This
information is fed into the manufacturing system
modelers.

Planner From the resoufce requirement and processing

time the planner develops the routing and material
handling requirements. This module considers the
Manufacturing various manufacturing system constraints and determines
the processing sequence. Once this is determined the
I material handling requirements can be finalized.

c
Tooling,
Currently human experts are required at this stage also.
fixture design
However, research to automate this stage are being
conducted elsewhere and this module will be updated as
[PRODUCTION j developmeats are reported.

Manufacturing System Models The information from

master scheduler and planner modules is fed into the
Figure 1 mathematical models of the manufacturing system. Two
System Design for Research Question I types of models are to be used: Static Model and

237
Dynamic Model. The models output various measures of that unduly affect the manufacturing system performance.
the manufacturing system performance. At this stage in When such a recommendation is made in the next product
the system we have created a link between manufacturing development meeting the designer may not be in
system performance and product design. agreement. The designer may want certain manufacturing
Static Modeling The input to these models consist of system parameters changed to accommodate this feature.
demand, resource requirements, and process flow. The In the original system such trade-offs and decisions were
performance measures output from these models consist to be made on an ad hoc basis. In the modified system a
of WIP, scrap, machine utilization, flow time and multi-attribute utility module will help alleviate the
throughput. This modeling technique is also called rapid situation. Figure 2 shows the modified system design
modeling. An example of a commercially available rapid required to address these situations. During system
modeling software is MANUPLAN [15]. These models development interaction with enterprises will be required
are typically not data intensive and are used in the initial to determine certain utility module parameters.
stages of the manufacturing system performance The modified system design can be obtained
evaluation. The models output only average values of the from the original system design with the addition of the
performance measures. This model is ideally suited to utility module. This system functions along the same
study the changes in specific system performance lines as the original system. The output from the output
measures due to changes in input parameters. analysis module is transferred electronically to the utility
Dynamic Modeling These modeling tools are more module. There the utility value for that product is
popular than the static modeling tools. Event-based computed. If the utility value is not satisfactory then a
simulation software packages fall in this category, decision has to be made whether to alter the
example SIMAN [16]. The data requirements are more manufacturing system parameters or the product design
intensive than for static modeling. However, the output parameters. The system will give the experts the data
measures are also more accurate and can be observed in needed to make this decision. Once this decision is made
system time. The model is ideal for detecting the worst the features or parameter to be changed is changed, and
case and best case scenarios. An expert will be able the process is repeated until a satisfactory utility value is
combine the information contained in both the models and obtained.
benefit from the rich mix of data that the models offer.
Utility Module In this module the goodness of product
Output Analysis The output of the models is analyzed. design and manufacturing system performance is
In this module individual design features that, if evaluated together. The product designer selects the
modified, would lead to better system performance are design attributes critical for the product, for example,
identified. The following methodology is proposed to be surface finish, paint luster, weight, etc. The plant
used to identify the features to modify: manager selects his critical manufacturing system
- identify the critical performance measure performance measures, for example, flow time,
- identify the cause of the problem throughput, WIP, queue length at the milling
- locate the resource@)where the problem initiates workstation, etc. The selection of these parameters is
- identify the feature(s) that are processed at the influenced in part by the other products to be processed in
resource@) the same manufacturing system at that time.
Once the features are identified a decision has to The utility module to be used in this system is
be made, whether to change the features. If a decision is similar to the one used by " t o n and Camahan [14].
made to change the features then this information is fed In the initial stages there is a uncertainty regarding the
back to the CAD system and the feature modified. estimates of the performance measures. If these
Developing standadized techniques that identifies the uncertainties are not included in the utility module they
design features to be changed will be one of the focuses will have a detrimental impact in the evaluation of design
of the research. alternatives. These uncertainties can be accommodated
by modifying the utility equation as follows:
Manufacturing Suppot? If the design is satisfactory
manufacturing and support make the necessary toolig
and fixture design required to manufacture &e product.
Then the product is ready for production.
where
When the system described in the previous
section is in place several potential pr&iem situations
may arise. For example, from the & p n t malysis module E[Uj] = '7Uj(x,)f(xj)dxj
of the plant manager may identify cer(ain surface finlshes xdm

238
 f(xj) is the probability density functions associated with
single attribute utility values, x- and xmax are the
minimum and maximum perfonnance measures for
attribute j, K and kj are as described before. These
I features. values will different based the characteristics of
enterprises. E[U] is the expected overall utility of the
1 I I design. The individual enterprises can set minimum
levels for these values before a design is accepted. The
utility module will also perform a sensitivity analysis that
database estimates which attribute has the most effect on the
overall utility value. This helps determine the parameter
ESouTce
requirements,
or design feature to change.
processing times
4. EXPECTED RESULTS AND BENEFITS
When the initial system is implemented the
capability to adapt product design for
manufacturing system performance will be
realized.
MANUFACTURING SYSTEM When the modified system is also installed the
MODEL users will also have the capability to ensure that
their entire design and manufacturing system has
been geared to provide them with maximum
WIP, scrap, dynamic queue utility.
machine changes, dynamic The greatest potential beneficiaries are make-to-
utilization, machine utilization, order companies (also called job shops), because
7 demand satisfaction f l o w h e . WIp these companies produce many different products
at one time and the product mix they produce
also c b g e s frquently.
Because of lack of previous research in this area
it is difficult to estimate the quantitative cost
manufacturing benefits that will accrue when the system is used.
parameters
Since, manufacturing cost forms a major portion
of the life cycle cost of a product even small
savings in this area would have a positive effect
on the cost of a product.
One of the salient features of the proposed
system is that significant benefits can be realized
quickly by integrating existing technologies.
However, in order to fully realize. the benefits all
the proposed modules will have to be developed
and integrated.
f Support
Where other similar CE tools were implemented
it resulted in shorter product concept to market
time, and this was achieved at significant cost
Tooling, savings. Similar benefits can be expected from
fixture design the proposed system.
All the products being produced in the enterprise
I PRODUCTION I will benefit from improved manufacturing
system performance. This will lead to
significant market advantage for the enterprise.
Figure 2. Modified System Design With the implementation of the suggested system
for Research Question II
a theoretical link between design and
manufacturing system performance can be
derived, this has not been done thus far.

239
 REFERENCES Market: Part II", Industrial Eneineering, v23,
August 1991,pp 42-46.
1. "Manufacturing Systems Strategic Plan",
Report of the Manufacturing Systems 13. Srinivasa K. Bhat and David K. Beagan;
Committee. March, 1993. "Feature Based Data Management",
Mechanical Engineering, March 1989,pp 68-72.
2. Bin Wu; "ManufacturingSystems Design and
Analysis", Chapman and Hall, New York, 1992. 14. Deborah L. Thurston and James V. Carnahan;
"Intelligent Evaluation of Designs for
3. Richard B. Chase and Nicholas J. Aquilano; Manufacturing Costs", Concurrent Engineering
"Production Management and Operations Automation. Tools. and Techniaues, ed. Andrew
Management",sixth edition, IRWIN, IL 60430, Kusiak, John Wiley and Sons, INC. New York,
1992. 1992,pp 437-462.

4. Steven Ashley; "DARPA Initiative in 15. "MANUPLAN", Network Dynamics, Inc,

Concurrent Engineering", Mechanical Burlington, MA, 01803.
Engineering, v114, April 1992,pp 54-57.
16. "SIMAN",Systems Modeling Corp., Sewickely,
5. "Concurrent Engineering", IEEE SDectrum, PA 15143.
~ 2 8July
, 1991,pp 22-37.
17. Tien-Chien Chang; "Expert Process Planning
6. "ConcurrentEngineering Automation, Tools, for Manufacturing", Addison-Wesley
and Techniques", ed. Andrew Kusiak, John Publishing Company, Reading MA, 1990.
Wiley and Sons, INC. New York, 1992.

7. Johnathan S. Colton; "An intelligent Design for

Manufacturing System", Concurrent
Engineering Automation. Tools, and
Techniaues, ed. Andrew Kusiak, John Wiley and
Sons, INC. New York, 1992,pp 153-176.

8. Zuomin Dong; "Design for Automated

Manufacturing", Concurrent EnPineering
Automation. Tools. and Techniaueg, ed. Andrew
Kusiak, John Wiley and Sons, INC. New York,
1992,pp 207-234.

9. "Concurrent Design of Roducts and

Processes: A Strategy for Next Generation in
Manufacturing", ed. James L. Nevins and
Daniel E. Whitney,McGraw-Hill Publishing
Company, New York, 1989.

10. G. Boothroyd and P. Dewhurst; "Product

Design for Manufacturing and Assembly",
Msnufachtrinn Eneineering, April 1988, pp 42-
46.

11. John I. Izuchukwu; "IntelligentFoundation for

Rodtact Design Reduces Costs, TimeTo-
Market: Part I", Industrial Engineering, v23,
July 1991, pp 29-34.

12. John I. ERlchukwu, " M i g e n t Foundation for

PrOckKt D a s i Redoas Costs, TimeTo-

240