Modular Design for Mechanical Product Customization Over the Internet 1

1058-8337/01/$.50© 2001 Journal of Design and Manufacturing Automation 4(1):37–46 (2001)

Functionality-Based Modular Designfor Mechanical Product Customization

Over the InternetYan WangDepartment of Industrial Engineering, University of Pittsburgh, Pittsburgh, PA15261

Bartholomew O. Nnaji*

Department of Industrial Engineering, University of Pittsburgh, Pittsburgh, PA15261

Received: February 20, 2001; Accepted: May 2, 2001.

Abstract: There is now greater demand for product customization. For large-scale mechanicalproduct customization, there is great need to develop new methods that will ensure shorter designcycle, shorter time to market, reduced life cycle cost, and higher product quality. The emergence ofthe Internet makes mass customization possible. It is essential that tools that will support customerparticipation in product design and realization be developed. The critical tool is an easy-to-useComputer-Aided Design (CAD) system over the Internet where customers’ preferences can becaptured by functionality-based formalism. A new conceptual design methodology, functionality-based modular design, is presented in this article with the capability of high flexibility and speed,which supports mechanical product customization. A model is developed for modularization toaccommodate different system behavior requirements from users of CAD. An XML-compatiblelanguage, PML, which has good properties of interoperability, scalability, and extensibility, isemployed for product representation. This modular conceptual design approach is implemented in anew design platform.

Key Words: Conceptual Design • Customization • Modularization • Functionality • Computer-Aided Design •Computer Integrated Manufacturing • Rule-Based Reasoning • Case-Based Reasoning • Knowledge-BasedDesign • PML • XML

I. INTRODUCTION

The emergence of Internet technologies is having greatimpact on manufacturing industries. The world-widecomputer network has led to a considerable shorteningof time and space and has also led to significant changesin the traditional structure of supply chain manage-ment. This revolution is also facilitating the evolutionof “paperless” manufacturing, which has been the en-during dream of Computer-Integrated Manufacturing(CIM). Customers will be able to participate in product

design activities at different stages over Internet with-out the restriction of time and space.

Normally, in the development of new product, thedesigner is presented with specifications that essen-tially are preferences and constraints of the customer.However, there is no existing mechanism to transformthese conceptual specifications into form. This is largelydue to the complexity of the design problem that re-quires capturing design intents and multidisciplinaryconstraints. Currently, the typical way to help custom-ers to reorganize the functional requirements along

* Corresponding author.

2 Wang and Nnaji

with other requirements (cost, manufacturability, etc.)comprehensively is the direct consultation of domainexperts. The internet can serve as a new medium forthe communication among customers, designers, manu-facturers, and vendors. One of the challenges for prod-uct customization over the Internet is the process ofcapturing customers’ intentions, requirements, andspecifications on a new product, especially for thosecustomers who may not have engineering design expe-rience. To understand customers’ requirements andaccommodate them in the design process, good Com-puter-Aided Design (CAD) systems, which should behuman-friendly, easy-to-use, intelligent, and experi-enced assistants, are crucial. In this article, functional-ity is perceived as the hub of design. A method offunctionality-based modular design is presented thatadvocates modularization and knowledge reuse in me-chanical design. This methodology is implemented ina new design platform.

II. THEORIES OF DESIGN

Considerable research effort has been spent on generaltheories about product design. At different levels ofabstraction, there are different categories. Accordingto Alexander,1 design is based on the idea that everydesign problem begins with an effort to achieve fitnessbetween two entities: the form in question and itscontext. The form is the solution to the problem; thecontext defines the problem. In other words, when wespeak of design, the real object of discussion is not theform alone, but the ensemble comprising the form andits context. Good fit is a desired property of this en-semble that relates to some particular division of theensemble into form and context. Freeman and Newell8

defined the design process as the process of transfor-mation of functional requirements into structure. Theywere among of the early researchers to introduce thenotion of functional reasoning as the basis for design.The Topological Design Theory24 is based on a topo-logical model of human intelligence. It starts withsome definitions (entity set, item of attribute, etc.).Based on some axioms (recognition, correspondence,and operation), it deduces theorems that modeldesigner’s designing process. The designing is the pro-cess of mapping between functional space and attribu-tive space. Suh18 defines design as the culmination ofsynthesized solutions in the form of product, software,processes, or system by the appropriate selection ofdesign parameters that satisfy perceived needs throughthe mapping from functional requirements in the func-tional domain to design parameters in the structuraldomain. The Domain Theory2 uses Function/Means-

Law and three domains for describing and synthesiz-ing a machine system. In the transformation domainthe purpose of any product or machine, which is tosupport a transformation or process, is captured. In theorgan domain, a materialized composition of products,which creates functions or effects, is the focus. In thepart domain, machine parts and their assembly rela-tions are considered. Design is a purposeful activity;4

it involves a conscious effort to arrive at a state ofaffairs in which certain characteristics are evident.Analysis, synthesis, and evaluation are three phases ofdesign.

Although researchers are using different models(e.g., Function-Attribute,24 Function-Behavior-State,20

Function-Behavior-Structure3,15), it is generally ac-cepted that design is the mapping process from func-tions to forms according to designers’ specifications orrequirements. Typically, these requirements will beexamined for functional, technical, operational, andfinancial feasibility by the assistantship of domainknowledge sources. For product customization, func-tionality is the starting point of the design and theultimate goal of a new product. Functionality is thecenter of the design activity. As shown in Figure 1, theoriginal functions should be considered in differentdesign stages, whether for form generation, materialselection, reliability analysis, or cost management.

Another reason that functionality plays an impor-tant role in product customization is that it is intuitivefor human beings, as new design is launched by thesense of not satisfying the existing ones. Customerscan utilize functionality as a channel to participate indesign so as to specify requirements, issue verifica-tions, and make decisions effectively with the com-puter.

III. CURRENT CAD SYSTEMS

Computer-Aided Design (CAD) software is playing animportant role for product design process. The earlyCAD systems just aided designers to finish drafting.Current CAD tools now include features that expressthe mechanical meanings of geometric configurationson a part.

The hub of the conventional mechanical CAD sys-tems is the geometric modeler. The task of the computeris focused on capturing geometric information, whereasnongeometric technical information (e.g., material prop-erties, functional requirements, manufacturing methods,etc.) and administrative information (e.g., bill of mate-rials, process planning and scheduling, cost estimation,etc.) are mostly neglected. This brings up the followingproblems of current CAD systems:

Modular Design for Mechanical Product Customization Over the Internet 3

1. Current CAD systems increase the risks ofdowngrading synthesis during analysis. Me-chanical designs need to consider materialproperties, tools selection, manufacturing, as-sembly processes, etc. They are much morecomplicated than pure electrical product de-signs from this perspective. The major reasonfor this is the difficulty of functionalmodularization and discretization of mechani-cal parts. Within electrical design, small mod-ules such as resistors and capacitors are thelowest level to consider for designers. Com-puter-Aided Drafting is not enough for me-chanical design. Integrated design tools ca-pable of reasoning about materials, processplanning, cost estimation, and other relatedissues are needed.

2. Current CAD systems are not “intelligent”enough to support error checking. Designersmay violate rules of geometry, ergonomics,

manufacturing, or assembly without aware-ness. For example, in a door lock design, anumber of errors could be made: a danglingedge is created; the handle for human users istoo short and edges are not rounded; or screwholes are too close to edge thus punch processmay disrupt the edge. An intelligent system isneeded to aid users and check these errors.

3. Current CAD systems lack mechanisms for“amateur” participation in a design process. Incurrent CAD systems, basic geometric infor-mation of a part such as geometric shape,dimensions, and features is specified by de-signers directly and explicitly. Thus, this islargely dependent on engineering knowledgeof designers. For customers, they usually lackthis knowledge. Their requirement may justbe “I need a structure here to place the shaft Ijust created” instead of “a º inch NPT threadedhole at the center of the frame”, or simply “I

FIGURE 1. Function as the hub of design.

4 Wang and Nnaji

need a tool to open a can”. A CAD system thatcannot follow customers’ intentions will notestablish effective human-machine communi-cation channels. Customers’ specifications tendto be ambiguous.

Research on advancing of the conventional CADparadigm, namely, Intelligent Computer-Aided De-sign (ICAD), has been on-going in the last 20 years. Ingeneral, ICAD aims at more intelligent use of comput-ers to aid in design process.

A group of researchers19,21,22 developed an IDDL(Integrated Data Description Language) to code de-sign knowledge in the IIICAD (Intelligent, Integrated,and Interactive Computer-Aided Design) system. Ituses the logic programming paradigm to express thedesign process for manipulating design objects, whereasthe object oriented programming paradigm is used toexpress design objects. The PERSPECT system5 is adesign tool that aims to support the effective utiliza-tion of experiential knowledge in numerical engineer-ing design, in which design knowledge is generated bylearning about their existing design domain (DomainExploration) and “sharing” the learning activity be-tween designers and computers (Shared Learning) soas to ensure the knowledge represented in computers isunderstandable to designers. CASECAD10,11 is a mul-timedia case-based design system that integrates tradi-tional CAD and case-based reasoning. It stores andutilizes design cases in both textual and graphicalmodes. The main modules include case memory, casebase manager, case-based reasoner, CAD package, andgraphical user interface. The system IDEAL3 usesanalogical reasoning to retrieve knowledge of a famil-iar problem or situation that is relevant to a givenproblem and transfers that knowledge to solve thecurrent problem, which is cross-domain case-basedreasoning. None of these systems are able to provideseamless transition from functionality to form.

IV. DESIGN MODULARITY

The market competition demands shorter design cycle fora new product. In order to reduce time, cost and risk fornew product development, there is a need for new designparadigm based on functionality and modularization.Modular design has been adopted in a number of indus-tries, such as electronic products and software develop-ment. Each module has certain functions and is self-contained, that is, each module’s functions do not dependon other ones. The designers need not design productfrom scratch each time. This approach reduces designtime and encourages design automation.

A number of researchers have attempted to imple-ment design modularization for mechanical products.Erixon7 presents a Design for Modularity technique —Modular Function Deployment (MFD), which consistsof five major steps: Quality Function Deployment(QFD) analysis; modular concept generation; ModuleIndication Matrix (MIM) to identify possible modules;Modularity Evaluation Charts (MEC) for thoroughevaluation; Module Indication Matrix to identify op-portunities for manufacture and assembly.

Similarly, ‘Design Reuse’ aims to maximize theuse of this concentration (of engineering creativity andexpertise in design) by the reuse of successful pastdesigns in part and in whole for new designs.17 TheEngineering Design Conference ’986 with design reuseas its main theme was held in UK, June 1998. Researchon focused innovation, cognitive studies, computa-tional perspective, use of standard components, designreuse tools, methods, etc. were presented.

Rather than product or part modularization, whichis directly beneficial to designers and manufacturers asa phase of the whole production process, we focus ondesign modularization from CAD’s viewpoint, whichemphasizes the methodology for easy-to-use CAD.We are proposing a method of functionality-basedmodular design for mechanical products. This approachintends to achieve functional modularization of designobject so that designers can interact with computerscomfortably and computers can aid designers to com-plete design effectively.

The advantages of design modularization are

1. Reduce the time of design. Instead of designingeach product from basic geometric elements,modules are taken based on their functions.Design engineers can reuse the modules with-out changes or change some sections of themodules according to the new product require-ments. The design history generated for rede-signing the unchanged sections will be saved.

2. Reduce the costs of design. Besides the timespent on design, the resources for productanalysis and evaluation are also reduced.Modularity helps to avoid errors and uncer-tainty during the design process.

3. Knowledge reuse. The design knowledge andexperiences are stored physically as part ofmodules. The risk of defective design is re-duced. The nature of design is to use old knowl-edge and acquire new knowledge about therelationship between functional space and formspace. Knowledge accumulation and storageis critical in design activities.

Modular Design for Mechanical Product Customization Over the Internet 5

4. Encourage concurrent engineering. The mod-ule is the media of communication among cus-tomers, market analysts, design engineers,manufacturing engineers, and other participantsof design. A common language based on mod-ule is beneficial for understanding each other.

The only possible pitfall in applying designmodularization is the possibility that a designer reliessolely on modules to develop new products. Anotherpitfall could be when modules are not well configured.Accurate module or accurate knowledge about moduleis the premise of the system.

Modularity and flexibility are two conflicting as-pects when modular design systems are developed.Modularity is helpful for routine design, while flex-ibility is essential for creative design. Both sides shouldbe considered in a practical CAD system. It is an art toachieve a good balance between those two aspects.One compromise is to use hierarchical modularity anddefine lower level modules if necessary to maintainsufficient flexibility. We define modus to be the small-est module that is an independent functional unit thatcontains one or more features.

Functionality-based modular design is a promisingmethod that facilitates Internet-based mechanical designand customization. Historically, the emergence of fea-ture-based design has changed the use of lines and curvesin mechanical design (which have no engineering mean-ing) to features that have an engineering implication. Thisnotion increased the abstract level of design reasoningand somewhat simplified the work of mechanical designengineers. As illustrated in Figure 2, early designers usedsimple geometric shape to model the product from view1. The emergence of features in design is showing in view2. The same product can be modeled in different ways.The functionality-based modular design intends to helpdesigners to model product from functionality viewpoint,as shown in view 3. Functionality is the focus of designfor supply chain participants as well as customers. Func-tionality module as defined by modus will be the com-monly used expression during a design process. The goalis to find a path such that design language and methodsare meaningful for customers apart from engineeringimplication. It is clear that functionality-based designallows designers to deal directly with the intentions ofdesign rather than the current indirect approach.

V. FORM GENERATOR INPEGASUS DESIGNER SYSTEM

Pegasus designer system is a CAD system under de-velopment at the University of Pittsburgh, which aims

to assist supply chain participants and customers toparticipate in new product design over Internet. Thegoal is to develop a new type of distributed CADsystem that supports product customization. Ease ofuse is the crucial requirement for this new designparadigm. Pegasus system is the fusion of CAD tech-nology, Internet technology, and knowledge engineer-ing, which facilitates information flow among cus-tomer, designer, manufacturer, and other stakeholdersfor product design. One of the functionalities of thissystem is the ability to allow a customer to generatemechanical product from concept and specifications tophysical form, by the aid of knowledge engineeringscheme. This task is accomplished by Form Generatorof Pegasus system.

A. Functional Modular RepresentationAlthough different design theories mentioned beforehave stated that design process is the mapping processbetween functional space and physical space, few con-crete mapping methods have been explored. Relationalapproach14 searches for working principles to fulfillsubfunctions. New design for a specific functionalitycan be derived by selecting the combined workingprinciples and working structures. Case-Based Rea-soning (CBR) has been used in various design fields,such as architecture, software engineering, structureengineering, etc.12,13

The mapping process used here is not directlyfrom functionality to component, which is employedin other approaches.9,16 Rather, mapping is from func-tion to geometric module by the aid of knowledgebase. The hypothesis here is that any function can beimplemented by one or more features. We define modusas the basic design operation unit that embodies func-tionality instead of feature or basic geometric elementssuch as line, circle, etc. The smallest modus, whichincludes one or more features, implements one func-tion. The hierarchical relationship among feature,modus, and component is shown in Figure 3.

Functionality is important in conceptual designstage, as it is the starting point to define a new design.In a top-down approach, conceptual or preliminarydesign is the process of functional decomposition. Wedisintegrate the functionality of products into Funda-mental Functionality Elements (FFE) — for example,support, hold, transport, convey, conduct, etc. It shouldbe realized that these FFEs are domain specific. Dif-ferent types of products have their own terminology.Defining FFE enables the system to define functionalmodule during design analysis and synthesis.

The element in functional space is FFE, while inphysical space the element is modus. The relationship

6 Wang and Nnaji

between FFE and modus is many-to-many, which isshown in Figure 4. A Fundamental Modus (FM) canbe defined as a modus that is only related to oneFundamental Functionality Element, while a Compos-ite Modus (CM) is related to more than one Funda-mental Functionality Element. In the example of Fig-ure 4, moduses m1, m3, m4, m5, are FMs and modusm2 is a CM. Because design is a particular activity ofhuman being, human intelligence and knowledge con-tribute significantly during this process. The mappingfrom functionality to form will proceed appropriatelyunder the guidance of knowledge base. These pre-ferred shapes or forms are chosen according to pastcases, experience and/or knowledge. The advantage ofdefining modus in the mapping is to increase the flex-ibility of conceptual design, comparing to component-based mapping.

Formally, we define F as the set of functionalities.f is an element of F (f ∈ F). Define M as the set ofmoduses. m is an element of M (m ∈ Μ). The mappingfrom functionality to modus is defined as p : F → M,which is the process of modus retrieval. The mappingfrom modus to functionality is defined as ′′ →→p M F: ,

which is the process of functionality retrieval. Thesemappings are not one-to-one because of their inherent

problem-dependent properties. Further, we define C asthe set of contexts, which captures the applicationenvironment of the target product. c is an element of C(c ∈ C). Then a more restricted mapping that tends toreduce the search space is p F C Mc: ×× →→ . Similarly,

′′ →→ ××p M F Cc: .

An abstract model of modus is built as follows:modus m can be defined as m S C G R SF O== , , , , ,

where SF is the set of functionalities that modus m isrelated to; C is the context that modus m is associatedwith; G is the geometric shape that modus m has; R isthe material of which modus m is made; SO is the setof other information that modus m has, which makesthis model extendable.

The important relationships among moduses in-clude functional relationship and spatial relationship.The functional relationship of moduses is the rela-tionship of functionality sets associated with moduses.They have relationships of union (∪), intersection(∩), difference (\), etc. The spatial relationship is therelationship among geometric entities. We define aWorld Coordinate System (WCS) located at the ori-gin point of 3-dimensional Euclidean space. WCS isthe reference of the whole product. The geometric

FIGURE 2. Different points of view for product design.

Modular Design for Mechanical Product Customization Over the Internet 7

entity of each modus has its own coordinate system,called Modus Coordinate System (MCS). The spatialrelationship among moduses is expressed in terms ofspatial relationship of MCS’s that moduses are lo-cated in. For convenience of expression, we define asymbol of reference ( TR ), in which TR is the trans-

formation of coordinate systems. A transformation isthe combination of a series of translations and rota-tions. TR is the transformation that one MCS or WCSneeds to finish to get the new MCS. For example,G TR G1 2 means geometric entities of modus 2, G2,

is positioned in MCS2 that is referring to MCS1 ofgeometric entities of modus 1, G1, after a coordinatetransformation.

The flexibility of the modularity is enhanced by aspecial binary operation, composition (⊕), betweentwo moduses. For two moduses,m S C G R SF O1

11 1 1

1== , , , , , m S C G R SF O22

2 2 22== , , , , ,

define:

m m m1 2 3⊕⊕ == ,

where m S S C G TR G R SF F O31 2

1 1 1 2 11== ∪∪ , , , , .

m m m2 1 4⊕⊕ == ,

where m S S C G TR G R SF F O41 2

2 2 2 1 22== ∪∪ , , , , .

Along with a composition, a transformation shouldbe defined such that the relative positions of twomoduses can be determined in 3D space. An exampleof composition is shown in Figure 5. This operation isnot permutable, i.e., m m m m1 2 2 1⊕⊕ ≠≠ ⊕⊕ .

⊕ can be looked as an operation of function andform union in the set of moduses. The result of theoperation composition is a composite modus that hasthe functional and geometric information of both modusoperands. After geometric entities of moduses are gen-erated and necessary compositions are finished, eachmodus can then be working as an independent compo-nent. Composition combines two moduses, thus re-duces the number of components. The design process

FIGURE 3. Relationship among feature, modus, and component.

8 Wang and Nnaji

FIGURE 4. Mapping from functional space to physical modus space.

FIGURE 5. Modus composition operation.

Modular Design for Mechanical Product Customization Over the Internet 9

can then be looked as the process of functional decom-position in functional space and modus composition inmodus space. The structure of a product can be graphi-cally modeled by a hypergraph, which is shown inFigure 6.

B. Functional ReasoningThe modus retrieval based on functionality, context,and other specification is accomplished by the aid of areasoning engine. Both rules and cases are necessaryfor the reasoning. Scientific rules, such as conserva-tion of energy, principle of hydraulics, etc., are impor-tant at the basic and abstract stage of reasoning. Engi-neering principles then are followed when reasoningcomes into a more specific area.

A reasoning procedure example during a centrifu-gal pump design is shown in Figure 7. When thedesign process starts, the designer system asks the userthe principle functional requirement of the target newproduct. According to scientific rules, the reasoningengine deduces that this device should be able to trans-fer kinematic energy of water to potential energy. Thena pump comes up because the knowledge base has therecord that pump will accomplish the required job.There are different kinds of pumps available. Afteracquiring the parameters of working environment ofthe target pump, the system is able to suggest a cen-

trifugal pump to customers according to the rules ofpump type selection. Then the system can help users tofinish the functional decomposition and choose a se-ries of Fundamental Functionality Elements neededduring component-level analysis, according to bothscientific rules and engineering rules. The modus re-trieval is the following procedure that selects moduses,either Fundamental Modus or Composite Modus, whichcontain geometric, material and other information. Atthe modus level, users give detailed specifications tomoduses, such as dimensions, materials, etc. Eachmodus then can become a separate component. If nec-essary, composition operations are conducted to re-duce the number of components. In this example,blades, which propel water, and the impeller hub, whichsupports blades, can be integrated into one component.After all components are finished, a product assemblymodel can be built using spatial relationships amongcomponents.

The whole design process is aided by rule-basedreasoning and case-based reasoning. Rules are appliedfor helping users’ decision making. Cases are stored atthe level of modus in the system.

C. Modeling LanguageHistorically, CAD format has presented a problematicbarrier for CAD information exchange. To encourage

FIGURE 6. Hypergraph model of product.

10 Wang and Nnaji

FIGURE 7. Flow diagram of functional reasoning in a centrifugal pump design.

Modular Design for Mechanical Product Customization Over the Internet 11

future information flow and CAD application overInternet, a system independent data format is necessaryand vital. With the emergence of Extensible MarkupLanguage (XML),25 the data exchange over Internet canhave a uniform format. XML is a simple, flexible, andstructured text format derived from Standard General-ized Markup Language (SGML) (ISO8879). Originallydesigned to meet the challenges of large-scale electronicpublishing, XML is playing an increasingly importantrole in the exchange of wide varieties of data over theInternet. It is becoming the backbone of Business-To-Business (B2B) electronic information exchange. AnXML-compatible modeling language, Product MarkupLanguage (PML), is developed in Pegasus. It is ex-pected to result in a new CAD system-independentmodeling language for mechanical product design.

PML has the following characteristics: (1) Exten-sible: By the nature of markup, PML can be extendedfor new information if necessary. When new conceptsor notations are to be used, the modeler can extend itslanguage scope by adding new elements in. It providesgood scalability for the modeler. (2) Portable andinteroperable: The language tends to separate system-independent content and system-dependent format ofproduct information so that useful information aboutproduct will not be lost during data exchange and trans-lation. PML can include more information in productfiles. It has the capability to include technical informa-tion, such as materials, tools selection, cutting path, etc.,and managerial information, such as order number, cost,etc., as well as geometric information from differentlevels. (3) Object-oriented: The inherent hierarchicaltree structure of the language enables good encapsula-tion so that modular transparency is guaranteed for thetop-down approach of design. Products are modeled byPML, which describes the information about the prod-uct explicitly, such as geometries, functions, features,materials and contexts, etc. Theoretically all informa-tion about product can be modeled in PML. (4) Compat-ible to Information Infrastructure: XML is looked as thefuture of web technology. PML is compatible to theWeb standard. Compatibility is indispensable whenbuilding an open system. Figure 8 shows the formerpump example modeled in PML.

D. Form Generator ConfigurationAs a part of Pegasus system, Form Generator accom-plishes the tasks of form generation from concept duringconceptual design stage. To accommodate different lev-els of users, modular design is introduced in Generator.

As shown in Figure 9, Generator is comprised ofGraphic User Interface, Modeler, Expert Shell, and associ-ated libraries and knowledge base. Because moduses are

the elements that users are dealing with, Generator tends tobe built as a flexible system that allows extension, that is,users are able to define their own moduses based on fun-damental moduses and available geometries, materials,and other information. Also these libraries and knowledgebase allow users to extend and update the information ifrequired. Within the system, Expert Shell plays an impor-tant role to aid users to finish design based on logic and itsknowledge. Modeler enables users to complete modusoperation and product information integration. The userinterface of Form Generator is shown in Figure 10.

VI. FUTURE WORK

Pegasus system is a pioneer web-based collaborativedesigner system. The ultimate goal is to develop an easy-to-use CAD system that supports mass customization inthe future. The system will have the ability of finishingconceptual design, detailed design, design test and evalu-ation, performance simulation, and design optimizationin real-time and on-line. Form Generator is the first stepusers will encounter when entering the system. Based onfunctionality, modular design is one of the pragmatic andrapid ways to accomplish conceptual design automation,considering both flexibility and speed.

The future work of Form Generator will be extend-ing the capability of the system to support detailed design.Considering manufacturing, assembly, reliability, main-tainability, ergonomics, and other related aspects, thesystem should be able to impose multidisciplinary con-straints onto product, as well as customer’s preferences.Those constraints should be captured and represented inappropriate ways so that those information will be trans-ferred and be constructive for later stages of productdesign.

VII. CONCLUSION

A new conceptual design methodology, functionality-based modular design, is presented in this paper. It hasthe capability of high flexibility, fast speed in support-ing mechanical product design and customization. Amodel of modus is developed to attain modularizationto accommodate different system behavior require-ments from users of CAD over the Internet. An XMLcompatible language, PML, which has good propertiesof interoperability, scalability, compatibility and ex-tensibility, is used to model products. PML is able toencapsulate both geometric and nongeometric infor-mation in an attempt to solve the existing CAD fileformat problems. A module that supports functionalmodular conceptual design is built to implement theabove developments.

12 Wang and Nnaji

FIGURE 8. A pump in Product Markup Language.

Modular Design for Mechanical Product Customization Over the Internet 13

FIGURE 10. Graphic User Interface of Pegasus Form Generator.

FIGURE 9. Architecture of Pegasus Form Generator.

14 Wang and Nnaji

REFERENCES1. Alexander, C., Notes On The Synthesis of Form, Harvard

University Press, Cambridge, Massachusetts, 1964.2. Andreasen, M.M., Conceptual Design Capture. In Pro-

ceedings of Engineering Design Conference ’98 on De-sign Reuse, S. Sivaloganathan and T.M.M. Shahin (ed.),Professional Engineering Publishing Limited, London,1998.

3. Bhatta,S.R. and Goel,A.K., From Design Experiencesto Generic Mechanisms: Model-based Learning in Ana-logical Design, Artificial Intelligence for EngineeringDesign, Analysis and Manufacturing: AIEDAM, Vol.10,No.2 (April), pp.131–136, 1996.

4. Coyne, R.D., Rosenman, M.A., Radford, A.D.,Balachandran, M., and Gero, J.S., Knowledge-BasedDesign Systems, Addison-Welsey, Reading, Massachu-setts, 1990.

5. Duffy, S.M. and Duffy, A.H.B., Sharing the LearningActivity Using Intelligent CAD, Artificial Intelligencefor Engineering Design, Analysis and Manufacturing:AIEDAM, Vol.10, No. 2 (April), pp.83-100, 1996.

6. Sivaloganathan, S. and Shahin, T.M.M. (eds.) Proceed-ings of Engineering Design Conference ’98 on DesignReuse, Professional Engineering Publishing Limited,London, 1998.

7. Erixon, G., Design For Modularity, In Design For X,G.Q. Huang (ed.), pp. 356–379, Chapman & Hall, Lon-don, 1996.

8. Freeman, P. and Newell, A., A Model for FunctionalReasoning in Design, proc. 2nd IJCAI, London, pp.621–640, 1971.

9. Gorti, S.R. and Sriram, R.D., From Symbol to Form: aFramework for Conceptual Design, Computer-AidedDesign, Vol. 28, No.11, pp.853–870, 1996

10. Maher, M.L., Balachandran, M.B., and Zhang, D.M.,Case-Based Reasoning in Design, Lawrence ErlbaumAssociates, Mahwah, NJ, pp. 141–164, 1995

11. Maher, M.L., CASECAD and CADSYN: Implement-ing Case Retrieval and Case Adaptation, in Issues andApplication of Case-Based Reasoning in Design, M.L.Maher and P. Pu (eds.), Lawrence Erlbaum Associates,Mahwah, New Jersey, Chapter 7, pp.161–185, 1997.

12. Maher, M.L. and Garza, A.G.S., Case-Based Reason-ing in Design, IEEE Expert, Vol. 12, No. 2 (March-April), pp. 34–39, 1997

13. Maher, M.L. and Pu, P. (eds.), Issues and Applicationsof Case-Based Reasoning in Design, Lawrence ErlbaumAssociates, Mahwah, New Jersey, 1997.

14. Pahl, G. and Beitz, W., Engineering Design: A System-atic Approach, Springer-Verlag, London, 1996.

15. Qian, L. and Gero, J.S., Function-behavior-structurePaths and Their Role in Analogy-based Design, Artifi-cial Intelligence for Engineering Design, Analysis andManufacturing: AIEDAM, Vol. 10, No. 4 (September),pp. 289–312, 1996.

16. Rosenman, M. and Wang, F., CADOM: A ComponentAgent-based Design-Oriented Model for Collaborative

Design, Research in Engineering Design, Vol.11, pp.193–205, 1999.

17. Sivaloganathan, S. and Shahin, T.M.M., Design Reuse:An Overview, Proc. Instn. Mech. Engrs., Vol.213, PartB, pp. 641–654, 1999.

18. Suh, N.P., The Principles of Design, Oxford UniversityPress, Oxford, 1990.

19. Tomiyama, T., Object Oriented Programming Paradigmfor Intelligent CAD Systems, in Intelledge CAD SystemII: Implementational Issues, V. Akman, P.J.W. tenHagen and P.J. Veerkamp (eds.), pp. 3–16, Springer-Verlag, Berlin, 1989.

20. Umeda, Y., Ishii, M., Yoshioka, M., Shimomura, Y.,and Tomiyama, T., Supporting Conceptual Design Basedon the Function-Behavior-State Modeler, Artificial In-telligence for Engineering Design, Analysis and Manu-facturing: AIEDAM, Vol. 10, No. 4 (September), pp.275–288, 1996.

21. Veerkamp, P., Akman, V., Bernus, P. and ten Hagen,P., IDDL: A Language for Intelligent Interactive Inte-grated CAD Systems, in Intelledge CAD System II:Implementational Issues, V. Akman, P.J.W. ten Hagenand P.J. Veerkamp (eds.), pp. 58–74, Springer-Verlag,Berlin, 1989.

22. Veth, B., An Integrated Data Description Language forCoding Design Knowledge, in Intelledge CAD SystemI: Theoretical and Methodological Aspects, P.J.W. tenHagen and T. Tomiyama (eds.), pp. 295–313, Springer-Verlag, Berlin, 1987.

23. Yoshikawa, H., General Design Theory and a CADSystem, in Man-Machine Communication in CAD/CAM,North-Holland, Amsterdam, 1981.

24. Yashikawa, H., Introduction to General Design Theory,in Intelligent Manufacturing Systems I, V.R. MilaV

CiC(ed.), pp. 3–17, Elsevier Science Publishers B.V.,Amsterdam, 1988

25. WWW Consortium, http://www.w3.org/XML/

BIOGRAPHIES

Bart O. Nnaji has a B.S. in Physics from St. John’s Universitywith distinction; an M.S. and Ph.D. in Industrial and SystemsEngineering from Virginia Tech, and obtained a certificate ofPostdoctoral studies in Artificial Intelligence and Robotics fromMIT. He was Professor of Department of Mechanical andIndustrial Engineering at the University of Massachusetts atAmherst till 1996. He is currently the ALCOA FoundationProfessor in Manufacturing Engineering at the University ofPittsburgh. Professor Nnaji has received 3 honorary doctoratesfrom international universities. He received the 1998 Outstand-ing Young Manufacturing Engineer Award by the Society ofManufacturing Engineering; 1992 Outstanding Young Indus-trial Engineer Award; He is a Fellow of Nigerian Academy ofSciences; Fellow of the Institute of Industrial Engineers, andFellow of the Society of Manufacturing Engineers. He washonored wit the U.S. Secretary of State’s Distinguished PublicService Award in 1995; Distinguished Scientist Award by the

Modular Design for Mechanical Product Customization Over the Internet 15

World Bank — IMF in 1998; and the Nigerian President na-tional honor — Officer of the Order of Niger (OON) in 2000.Professor Nnaji has served as principal or coprincipal investiga-tor on over $35 million research. He has published 5 books andover 100 technical articles. One of his books, Computer Inte-grated Manufacturing and Engineering, won the 1994 worldbest text book prize for Manufacturing Engineering. ProfessorNnaji is the founding Editor-in-Chief of the International Jour-nal of Design and Manufacturing and also serves as the Editorfor the Design Department of Institute of Industrial Engineers

Transactions on Design and Manufacturing. Professor Nnajiserved as Nigeria’s Federal Minister of Science and Technologyin 1993.

Yan Wang is a Ph.D. student in Department of IndustrialEngineering, University of Pittsburgh. He received his MSEEdegree from Chinese Academy of Sciences in 1998, BSEEfrom Tsinghua University, China, in 1996. His current re-search interests include 3D geometric modeling and collabo-rative knowledge-based CAD.