A web-based fuzzy mass customization system

Journal of Manufacturing Systems

A Web-Based Fuzzy Mass Customization System Y.H. Chen,Y.Z. Wang, and M.H. Wong, Dept. of Mechanical

Engineering,

The University

of Hong Kong.

E-mail: yhch~n, yzwang~hkucc.hku.hk

Abstract

ments is a prerequisite for realizing mass customization. Companies must initiate a dialogue with individual customers to help them articulate their needs (Proops 1996). A number of techniques have been used to facilitate customer input into the design phase of product development. These include surveys, focus groups, and customer interviews. But these techniques do not capture the complete picture of customer preferences (Brown, Hitchcock, and Willard 1994). Users rather than suppliers are the actual designers of the application-specific portion of a product (Von Hippel 1998). The Internet has turbocharged companies’ abilities to track individual customer preference (Dewan, Jing, and Seidmann 2000). It is unsophisticated to construct a website wherein customers can simply select offered options from lists to generate different product or service combinations. However, in markets like automobiles or consumer goods, which have a crucial emotional connection between customer and product, there is a high emphasis on an appealing and aesthetic exterior of the product. Styling is often the final differentiation criterion among products competing on the market (Dankwort and Podehl 2000). More market share will be captured and higher profits gained if a firm can let mass household consumers customize the style of the products they want. However, due to the limited network bandwidth and the huge amount of data in the 3D geometry description, it is impossible to implement a full version of a CAD system on today’s Internet. On the other hand most commercial CAD systems currently in use are too sophisticated to beginners. The research work in this paper therefore concentrates on the development of a web-based application that allow both beginners and seasoned users of CAD systems to define and customize certain families of products on the World Wide Web.

Mass customization attempts to give customers the product they want, when and where they want it, at a cost comparable to that of mass-produced goods. With the increasing popularity of the Internet and broadband, it is now possible to let mass household consumers be involved in the design of a product that reflects their preferences or personalities. As such an application is targeted at the general public, a descriptive or linguistic input style is preferred. The concept of fuzzy customization is therefore proposed and investigated. A prototype system is implemented on a web clientkerver architecture, namely CyberFGC, which consists of a fuzzy geometric customization (FGC) program, Virtual Reality Modeling Language (VRML), and common gateway interface (CGI) programs. In this system, household consumers can customize products using their preferred linguistic description such as big, small, normal, etc., over the World Wide Web. Examples of customization of wine glasses and furniture are described in detail. Keywords: Mass Customization, Fuzzy Reasoning, VRML, Geometric Modeii~g

Introduction Due to increasingly demanding customers and market turbulence caused by the globalization of modern economy, the manufac~ng indus~y, and many other industries, has been expe~en~ing a fundamental change-from mass production to mass customization. Mass customization, coined by Davis (1987), aims at developing, producing, marketing, and delivering affordable goods and services with enough variety and customization that nearly everyone finds exactly what he/she wants (Pine 1993). Mass customization challenges traditional manufacturing precepts by using mass production processes to meet the singular needs of individual customers. To achieve the goals of mass customization, manufacturing companies must focus on their competitive advantage (Daflucas 1998). Mass customization starts with understanding customers’ individual requirements (Fulkerson 1997). Effective definition of customer require-

280

Journal of Manufacturing Systems Vol. 2a/No. 4 2001

With the widespread use of the Internet, webbased design is attracting more and more attention from the CAD/CAM community. The majority of recently reported research focuses on the development of certain methodologies that are applicable to effective communication and cooperation among the working groups that are dispersed all over the world. A substantial American research project-the ARPA Manufacturing Automation and Design Engineering (MADE)-was conducted from 1992 to 1996 (Cutkosky, Tenenbaum, and Glicksman 1996; Petri 1996). The MADE program develops Internet-based tools, services, protocols, and design methodologies for design and manufacturing teams. It is concerned with the comprehensive information modeling and the design tools needed to support rapid design of electromechanical systems. Gadh and Sonthi (1998) discussed the need of different levels of geometric abstractions to enable design teams collaborating over the Internet at different stages of the product development process. Roy et al. ( 1997) reported a web-based collaborative product development environment where the Web was used for the sharing of CAD data and product design data. Similar works are also reported by Huang, Huang, and Mak (2000); Kim, Lee, and Han (1999); Summers and Butler (1999); and Wanger, Castenotto, and Goldberg (1995). In the product development process, various attempts have been made. Huang, Huang, and Mak (1999) and Huang and Mak (1999) developed a generic, web-based Design for X (DFX) shell that can be tailored or extended to develop and apply a variety of DFX tools easily, quickly, and consistently. The CyberCut concept that provides the first design-to-manufacturing rapid prototyping system over the Web was reported recently (Smith and Wright 1996). The CyberCut concept is the synthesis of World Wide Web technology and interconnected CAD/CAPP/CAM components of the Integrated Manufacturing and Design Environment (IMADE) developed at the University of CaliforniaBerkeley. CyberCut constrains the user to the design of parts that are manufacturable on a three-axis milling machine. All of the above researches are mainly focused on conceptual development. Three-dimensional visualization on the Web is not available. In this paper, product families are implemented as parametric models. The solid model of a product is created in

such a way that the model creation operation can be re-executed with new values for the defining parameters of the associated geometric features (Cox 2000). Because the system is developed for the public at large, parameters that describe a family of products are defined as fuzzy variables. Fuzzy logic can handle various types of vagueness and uncertainty, particularly the vagueness related to human linguistics and thinking (Tanaka 1997). In the proposed system, a web server/client architecture, namely CyberFGC (Cyber Fuzzy Geometric Customization) is proposed and implemented. It consists of a fuzzy geometric customization (FGC) program, Virtual Reality Modeling Language (VRML), and common gateway interface (CGI) programs. FGC is used to accept customers’ inputs and generate design models. VRML, an IS0 standard language for describing 3D models on the Web, is used for 3D model visualization.

Fuzzy Reasoning Fuzzy reasoning is performed by inference rules, which are expressed in IF-THEN format. The linguistic rules can simulate human thinking processes to some extent. IF-THEN rules used in fuzzy reasoning are called “fuzzy IF-THEN rules,” which represent the relation (or transformation) of input variables to the output and are normally expressed as: IF Ai and/or Bi, THEN Hii, else IF A2 and/or B,, THEN Hz,, else IF A, and/or B2, THEN H12, else IF A2 and/or Bz, THEN Hz2 The defuzzification process is an important step in fuzzy reasoning. In general, defuzzification is the process where the membership functions are sampled to find the grade of membership; then the grade of membership(s) is used in the fuzzy logic equation(s) and an outcome region is defined. From this, the output is deduced. Several techniques have been developed to produce an output. Centroid, which takes the center of gravity of the output fuzzy set as output value (Tanaka 1997), is adopted in this research; that is:

281

Journal of Manufacturing Vol. 2oLNo. 4

Systems

2001

1

small

very small

normal

large

very large

~

/ 0

j

: 0.2

:

j 0.4

j

; 0.6

/

: 0.8

:

j 1.0

Figure 2 Membership Functions small

very small (a) Side view

Parameters

normal

large

very large

(b) Isometric view

Figure 1 of a Cocktail Wine Glass

~

:

y”=-Jj&!(Y)dY

;o

(1)

46

Membership

where J means ordinary integral.

( $2

: ;8

: :

j

124

150

Figure 3 Function of Fuzzy Variable A

shows the membership functions of the fuzzy sets, which are divided into five grades: very small, small, normal, large, and very large. “0” and “1” imply the minimum and maximum possible values, respectively. When a range is defined for a given fuzzy variable, such as the glass opening diameter A as in Figure I, the membership function is automatically mapped to the fuzzy variable. For instance, if A is defined as [20 mm, 150 mm], the membership function of A is as shown in Figure 3.

Fuzzy Geometric Customization Fuzzy geometric customization (FGC) is aimed at developing a new method of geometric customization that can be easily used by ordinary users. FGC features linguistic input and only requires the definition of important parameters. Using FGC, an instance of a parametric model is modeled by linguistic description. Due to the nature of linguistic description, FGC has the potential to attract mass household customers through the Internet.

Parameter Deduction The parameters of a product family are divided into two sets: user-defined parameter set U and deduced parameter set D. Normally, parameters that strongly influence the shape of a product, such as height, width, diameter, etc., are defined as set U that requires input by a user. The deduced parameter set D contains parameters that do not sharply influence the shape, such as fillet, chamfer, etc. The deduced parameter set D is determined by set U using fuzzy inference. This is convenient and timesaving because the exact mathematical model of a part is very complicated. In the cocktail glass shown in Figure I, parameters A, B, C, D, E, T,, and T2 are important ones that are defined as set U. A user can customize the geometry or the shape of a cocktail glass by changing one or some or all of the parameters. The fillets F,, F2,

Parameter Definition Many products fall into the class of design that may be regarded as parametric. Parametric designs have similar shape, only having variations in their geometric dimensions. For example, a cocktail wine glass as shown in Figure 1 can be defined by a set of parameters. The parameters are divided into two sets. The user-defined set U{A,B,C,D,E,T,,T2} input from users. The other set requires D{F,,F,,F,,F,} is deduced from user-defined set U. A parametric model requires the explicit definition of both U and D. Fuzzy Input All input parameters are represented as fuzzy numbers. They are expressed by fuzzy sets. Figure 2

282

Journal of Manufacturing Qstems Vol. 2omo. 4 2001

Table I Tabulation of Decomposed Fuzzy Rules for F2

y---_ c\$

I --.!_‘very small

vet
\Illustration

k\,F;

Figure 4 of Parameter F,, B, C, and D

_

Fz’ very small

very small small normal large very large

F3, and F4 are defined as set D that are deduced automatically from set U through fuzzy inference. As an example, F2 smoothes the edge between the upper body part and the stem part. Figure 4 shows the cross-section diagram of the upper body and the stem part, the upper body height, and the stem height of the cocktail glass, and the location of F2. The stem diameter D, the upper body height B, and the stem height C are fuzzy variables, and the fillet between the upper body part and the stem part F2 is a fuzzy output. F,’ is defined as an intermediate fuzzy variable. For simplicity, the decomposed fuzzy rules for F2 are tabulated in Table I; the rules are generated in accordance with the property of the cocktail wine glass and common sense knowledge. Similarly, fuzzy rules are defined for all parameters in set D.

very small very small ~~ small normal normal

very small very small normal+ large large

T.1”L:, small small normal normal

( small ~normal

small

normal

! 2::

normal normal

1 large I large

‘very large [ve~~Zge~very large large

1very large

very small veryGiaT iiiidl. small small small small normal ) normal I normal normal large ~ large large large very large , large Ivery large ~ very large very large

information available to an external program, and CGI can send the output of a program to a web browser that requests it. Most CGI programs do both of these things. CGI is the engine that enables you to extend HTML by creating applications that tie together multiple web pages, to interact with the server, and to process data retrieved using HTML forms. The advantages that CGI offers are: platform independence, language independence, and scalability. In the proposed system, all CGI programs are written in Perl. Perl, a high-level programming language with an eclectic heritage, seems to be, far and away, the most popular language for writing CGI scripts (Castro 1999, Colburn 1998). Per1 derives from the ubiquitous C programming language and to a lesser extent from sed, awk, the Unix shell, and at least a dozen other tools and languages. Per1 is a powerful textmanipulation tool and is easily moved from one platform to another. Although Per1 supports large, highly structured applications, it is also easy to write short, simple scripts that are quite powerful. VRML, an acronym for the Virtual Reality Modeling Language, is a language for describing multi-participant interactive simulations or virtual worlds networked via the Internet and hyperlinked with the World Wide Web (Ames, Nadeau, and Moreland 1997). VRML completely supports 3D models with polygonal rendered objects, materials, etc. VRML supports the hyperlink feature. Models developed using VRML can be linked to other documents on the Web, which can contain textual, numerical, or audio/visual data. The first version of VRML was released in early 1995. VRML 97, the latest version of VRML, became an IS0 (International Organization for Standardization)

CyberFGC Construction The Web is a large collection of clients and servers that support the HyperText Transfer Protocol (HTTP) on the Internet. The information is stored on servers, and clients make requests for the information they need. The server and client machines exchange messages in the HyperText Markup Language (HTML) format. Both HTTP and HTML are open standards and are implemented on a wide variety of platforms on the Internet. CGI, Perl, and VRML CGI, standing for common gateway interface, is a protocol (a way of doing things) to connect web servers to external applications, not a programming language (Castro 1999, Colburn 1998). Any script that sends or receives information from a server needs to follow the standards specified by CGI. CGI can do two things. It can gather information sent from a web browser to a web server and make the

283

Journal of Man~factacturing Systems Vol. 2a/No. 4 2001

User m

User WI

User

6

T

User

V R

The Internet Input parameters

VRML models

The Server

Figure 5 CyberFGC Architecture

standard late in 1997 (Stone 1999). All 3D visualizations of a product design in the proposed system are in VRML models. CyberFGC To demonstrate the fuzzy mass product customization concept on the Internet, a web client/server architecture, namely CyberFGC as shown in Figure 5, is proposed and implemented on an Alpha Workstation 600au. CyberFGC consists of a fuzzy geometric customization (FGC) program, Virtual Reality Modeling Language (VRML), and common gateway interface (CGI) programs. The FGC deals with customers’ inputs in both precise numerical data and fuzzy linguistic information and generates design models accordingly at server side. VRML is used at client side to visualize design models in 3D. The advantage offered by VRML models is that the visualization of customized products can be dynamically updated. All of the above elements are interfaced by CGI programs, which are developed using Perl.

Figure 6 Main Page for Customization

of Wine Glasses

four types of wine glasses, namely a cocktail wine glass, champagne wine glass, snifter wine glass, and user-defined wine glass as shown in Figure 6, can be customized with user-friendly fuzzy parametric input or numerical input. VRML is used to dynamically visualize 3D design models of wine glasses. Figure 7 shows the fuzzy logic input page of a cocktail wine glass. The server receives and validates the received parameters, generates the models of the designed wine glass, and delivers its VRML model to the requesting client. As VRML is used for 3D model visualization, an extra plug-in VRML viewer is necessary in the web browser. VRML viewers, typically Cosmo Player 2.1, are companion applications to standard web browsers for navigation

Website Construction For better illustration, different websites for fuzzy customization of the wine glass and the furniture are constructed. The Uniform Resource Locator (URL) of the fuzzy wine glass customization website is http://hkumea.hku.hk/yhchen/designlWG.html, and the website for furniture customization is http://papeKheha.net. For the wine glass design,

284

Journal of Manufacturing Systems Vol. 20/No. 4 2001

~ ' ~ o ~ .

:

:

.......

{............. ; ..............................................

s~

-::...............................................

DESIGN RESULT

Cocktail W'me Glass O¢~ig~l wa,~ 6 1 . . ~ r ~ i ~ , ~ - ~ ~e~~a~ ~ ~o~

Inll~t M~*te

............... .............

c P~eter ~r*put ~" Fum.O' Log,c ~np~t mp,~,e ~ , ~ e , * . . . . . . . . . . . . . . . . . . . . . . . . . Pray I~put IUpper Part| Op ~r~g DlameIe¢ [A]

Upper Body }telght [B] Uppe~ Body "~¢~.cka~e$v IT] ] F~'

Pa~ I

Input [ L ~ r

Base D~ame~er [E] very sm~ ................................. Base ~ "ck~egsIT2] ve~ tirol

31ass Stem H ew)it [C] 3las~ StemDt~'neter D]

t

Base Di~neter IF*] Base Tmcka*esaIT2]

Figure 7

Figure 8

Fuzzy Logic Input Page of Cocktail Wine Glass

VRML Model of a Cocktail Wine Glass with Its Fuzzy Parameters

DESIGN RESULT

Whle Glass Type; Cockaed

D~'i~M~t~d:F!~ mp,~.

. . . . . . .

Input P~eter~

Glass Stem He~ghl [C] (~a~s Sty1 Eh~,eter [D]

very shor~

Base Diameter {El

uery ~m~

:

very small

Figure 10 Prototype Wine Glasses Defined Using the Proposed System

Figure 9 The Changed VRML Model

and visualization. Figure 8 shows the VRML model of a cocktail wine glass generated according to the input shown in Figure 7. The user can always change the parameters. For example, if the container height B is changed from very short to very tall, the VRML model of the design is consequently changed as shown in Figure 9. Thus, the user can customize the wine glass model until the one desired is obtained. To view a physical prototype of the wine glass design, an STL file of the model can be generated. Figure 10 shows a few wine glass prototypes that are designed using the website and built by selective laser sintering (SLS) equipment. Another interesting mass customization is 3D furniture design and layout. Figure 11 shows the chair customization page. It can be seen that all parameters describing the size of a chair are defined as fuzzy variables. An instance of a chair design is

Figure I1 Page for Customization of a Chair

selected by simply choosing a value from the pulldown menu of each fuzzy variable as shown in Figure 12. Similarly, the website also provides customization of tables, cupboard, lamps, and wallpaper. Examples of table and cupboard customization are given in Figures 13 and 14, respectively. After the design of all components, they can be placed in a room so that the overall assembly can be visualized. Figure 15 shows that the placing of furniture components is accomplished on the plan view of the room. A 3D visualization of the room is constructed as a VRML model as shown in Figure 16.

285

Journal qf Manu[acturing Systems Vol. 20/No. 4 2001

[ ~

,

~

Figure 12

Figure 13

Bar Chair with Certain User Input

Table Customization Page

~

Figure 14 Cupboard Customization Page

Figure 15

Figure 16

Layout of All Furniture on Plan View of a Room

Final Assembly Constructed as VRML Model

Conclusion

divided into two types: user-defined parameters and deduced parameters. All parameters are defined as fuzzy variables. The user-defined parameters are input by a user. The deduced parameters are determined by the user-defined parameters using fuzzy reasoning. A prototype system has been implemented on the Internet to demonstrate the proposed fuzzy mass customization concept. Through examples of wine

This paper has presented a new design approach, namely fuzzy mass customization, which allows most household consumers, who are not familiar with both mechanical design and sophisticated CAD software, to customize some parameters of a product using preferred linguistic information such as small, normal, big, very big, and so on. A family of products is represented using a set of parameters that is

286

Journal of Manufacturing Systems Vol. 2a/No. 4 2001

glass and furniture design, it can be seen that the proposed system is effective for products of simple shape or when only a few critical parameters of a complex product are frequently customized. Acknowledgment This research is carried out with a CRCG grant from the University of Hong Kong. References Ames, A.L.; Nadeau, D.R.; and Moreland, J.L. (1997). VRML 2.0 Sourcebook. New York: John Wiley & Sons. Brown, M.G.; Hitchcock, D.E.; and Willard M.L. (1994). why TQM Fails and What to DoAbout It. Burr Ridge, IL: Irwin Professional Publishers. Castro, E. (1999). PERL and CGIfir the World Wide Web. Berkeley, CA: Peachpit Press. Colbum, R. (1998). Sams’ teach yourself CGI programming in a week. Indianapolis, IN: Sams.net Publishers. Cox, J.J. (2000). “‘Product templates’ a parametric approach to mass customization.” In CAD Tools and Algorithms for Product Design, P. Brunet, C. Hoffmann, and D. Roller, eds. Berlin: Springer-Verlag, ~~3-15. Cutkosky, M.R.; Tenenbaum, J.M.; and Glicksman, J. (1996). “Madefast: collaborative engineering over the Internet.” Communications of the ACM (~39, n6), ~~78-87. Daflucas, M. (1998).“Road to mass customization.” Industrial Computing (~17, n6), ~~16-19. Dankwort, C.W. and Podehl, G. (2000). “A new aesthetic design workflow-results from the European project FIORES.” In CAD Tools and Algorithms for Product Design, F’.Brunet, C. Hoffmann, and D. Roller, eds. Berlin: Springer-Verlag, pp 16-30. Davis, S.M. (1987). Future Perfect. Reading, MA: Addison-Wesley. Dewan, R.; Jing, B.; and Seidmann, A. (2000). “Adoption of Internet-based product customization and pricing strategies.” Proc. of Hawaii Int’l Conf. on System Sciences, Jan. 4-7,2000, ~~135. Fulkerson, B. (1997).“A response to dynamic change in the market place.” Decision Support Systems (~21, n3), ~~199-214. Gadh, R. and Sonthi, R. (1998). “Geometric shape abstractions for Internetbased virtual prototyping.” Computer-Aided Design (~30, n7), ~~473-486. Huang, C.Q.; Huang, J.; and Mak, K.L. (2000). “Agent-based workflow management in collaborative product development on the Internet.” Computer-Aided Design (~32, n2), ~~133-44. Huang, C.Q.; Lee, S.W.; and Mak, K.L. (1999). “Web-based product and process data modelling in concurrent ‘design for X’.” Robotics and Computer-Integrated Mfg. (v15), ~~53-63. Huang, C.Q. and Mak, K.L. (1999). “Design for manufacture and assembly on the Internet.” Computers in Industry (~38, nl), ~~17-30.

Kim, H.; Lee, J.Y.; and Han, S.B. (1999). “Process-centric distributed collaborative design based on the Web.” Paper No. Detc99/CIE-9081. Proc. of DETC’99 ASME Design Engg. Technical Conf., Sept. 1999, Las Vegas, NV Petrie, C.J. (1996). “Agent-based engineering, the web, and intelligence.” IEEE Expert (~11, n4), ~~24-29. Pine II, B.J. (1993). Mass Customization: The New Frontier in Business Competition. Cambridge, MA: Harvard Business School Press. Proops, S. (1996). “Mass customization. Stimulating the knowledgeable market.” IEE Colloquium (Digest) (~181, nlO), ppl/l-l/6. Roy, U.; Bharadwaj, B.; Kodhani, S.S.; and Cargian, M. (1997). “Product development in a collaborative design environment.” Concurrent Engg. Research and Application (v4), ~~347-365. Smith, C.S. and Wright, PK. (1996). “CyberCut: a World Wide Web based design-to-fabrication tool.” Journal of Manufacturing Sy.stems (~15, n6), ~~432-442. Stone, M. (1999). “Virtual Reality Modeling Language.” IEEE Computer Graphics andApplications (~19, n2), pp17. Summers, J.D. and Butler, A.C. (1999). “Development of a feature based design system using virtual reality.” Paper No. Detc99/CIE-9034. Proc. of DETC’99 ASME Design Engg. Technical Conf., Sept., 1999, Las Vegas, NV Tanaka, K. (1997). An Introduction to Fuzzy Logic for Practical Applications. Translated by T. Niimura. New York: Springer. Von Hippel, E. (1998). “Economics of product development by users: the impact of ‘sticky’ local information.“Mgmt. Science (~44, n4), ~~629-644. Wagner, R.; Castanotto, G.; and Goldberg, K. (1995). “DFX via the Internet.” SPIE (v2596), ~~192-195.

Authors’ Biographies Dr. Y.H. Chen is an associate professor in the Dept. of Mechanical Engineering, The University of Hong Kong, since 1993. Before joining The University of Hong Kong, Dr. Chen has worked at Motorola Electronics Pte, Ltd. (Singapore), Asia Matsushita Electronics Pte, Ltd. (Singapore), and Swire Technologies Pte, Ltd. (Hong Kong) as senior automation engineer, senior research engineer, and automation manager, respectively. Dr. Chen’s research interests include CAD/CAM, reverse engineering, and rapid prototyping. Dr. Y.Z. Wang is currently a research associate working with Dr. Y.H. Chen. Dr. Wang graduated in precision instrument engineering from the Dept. of Precision Instrument Engineering of Tianjin University, P.R. China. He received his bachelor’s, master’s, and doctoral degrees in 1980, 1993, and 1996, respectively. Dr. Wang’s research interests include Internet-based product development, reverse engineering, and dimensional measurement. Mr. M.H. Wong is a research assistant in the Dept. of Mechanical Engineering, The University of Hong Kong. He is interested in CAD/CAM and Internet computing.