increasingly competitive markets Cohen(1995) proposed using Master HouseQuality for planning product variety Suh(1990) viewed product variety asthe proper selection of design parameters that satisfy variant functional re-quirements Ulrich(1995)examined the relationships between product archtecture and product variety, component standardization, modularity, andproduct development Erens(1996)developed product variety under functional, technology, and physical domains Fujita and Ishii(1997)formulated thetask structure of product variety design, and Martin and Ishii(1996, 1997,2002) proposed DFV (Design for Variety), which is a series of methodologieswith quantifying indices for reducing the influence of product variety onproduct life-cycle cost, and thus helping design teams to develop decoupledproduct architectures These studies have established a basis for product varity management However, many investigations have agreed that the key toefficiently designing and delivering multiple products is developing a goodproduct architecture(Meyer 1993, Sawhneyadvantages of developing product architecture is that it enables a company tooffer two or more products that are highly differentiated yet share a substantial fraction of their components The collection of components shared by theseproducts is called a product platform(Ulrich&z Eppinger 2000)

Erens(1996)defined a product platform as"An architecture concept of compromising interface definitions and key-components, addressing a market and being a base forderiving different product families Robertson and Ulrich(1998)proposed amethod of balancing distinctiveness with commonality within product architecture through identifying the importance of various factors going into thistradeoff Fujita et al,(1998, 1999)utilized optimization techniques to identifthe optimum architecture of a module combination across products in a familyof aircraft Moreover, Yu et al ,(1998)defined product family architecturebased on customer needs by using the target value of product features for calculating probability distributions Additionally, Simpson, et al,(1999)used theProduct Platform Concept Exploration Method(PPCEM) to design a commonproduct platform This platform uses the market segmentation grid to helpidentify suitable scale factors of the platform that are"scaled"or"stretched"tosatisfy various requirementAlthough most studies focus on optimizing product structure, some studienoticed that investigating the physical arrangement and interactionamong components is the key for stable product architecture For example, thecomponent-based DSM(design structural matrix) method has been applied to

timize variant Product Design Based on Component Interaction graNote: Grayed cells indicate the inter-chunk interactionsTable 4

Incidence matrix after appropriate rearranging the orderLevelLevel 3imponentConstraint口 ModuleFigure 4 Hierarchical graph of component interaction

anufacturing the Future: Concepts, Technologies VisionsTablethat Carafe(part No 25) design is essential for maintainingcoffee temperature Therefore, the Carafe is redesigned to meet the requirement: the wall should be thickened and use heat insulation material, the shapederized and the top narrowed to reduce heat loss To identify the influenceof the new Carafe design, Fig

5(extracted from Fig 4)shows the incidence diagram of the Carafe In this figure, the design constraints the Carafe exports tohe sink nodes are listed below:(fixed)8 Plate module should fit the diameter of Carafe based(25, 21): The heaterThe Cup Bank should fit the diameter of Carafe body(changed)The Carafe Handle should fit the arc and weightd(25, 26): The Carafe Cover should fit the diameter of Carafe rim and the requested thermal condition(changedd(25, 27): The Filter Module should fit the Carafe height(changed)Carafe27}5ting plateTop cover modul○ Source of variant designFigure 5 Incidence diagram of carafe(component 25)

timize variant Product Design Based on Component Interaction grae constraint of d(25, 21)is fixed (represented as dotted line in Fig 5), andthus parts 20, 21 are left unchanged However, constraints d(25, 22), d(25, 24),d(25, 26), and d(25, 27) are changed (represented as solid lines in Fig 5)owingto the new carafe specification, resulting in the design of the Filter Module andCarafe Outfit Module having to be changed to match the altered conditions Inhe Carafe Outfit Module, the components are redesigned to fit the new Caafe However, the design change of the Filter Module must refer not onlyche Carafe, but also to other components that provide constraints on the filterModule, as shown in Fig 6

Thus in redesigning the Filter Module, the constraint from the Carafe becomes the source of variant design(represented assolid line in Fig 6), while the others are fixed constraints(represented as dottedines in Fig 6) listed below:d(11, 27): The Housing should fit the based(28, 30): The Filter Holder should fit the Filter Holder packing valved(4, 5): The Top Cover Base should fit the Spout Seat diameterd(3, 30): The Filter Holder should fit the Spout shapeUnder these constraints, the design of Filter Module (parts 27, 5, and 30)changed from V-shaped to U-shaped to fit the new Carafe desiFurthermore, constraint from the filter module isd(5, 1): The Top Cover should fit the Basket Holder rim diameterSince the specification of the Basket Holder rim is fixed, component 1 and 2need not change their desionsequently, Table 5 lists the dedriven by spatial market differentiation0 Redesigned componentCup bank of theof thermaldle of thermal carafe26* Cover of thermal carafe7 U-shaped housingU-shaped cover baseTable 5 List of variant components foket 2

Manufacturing the Future: Concepts, Technologies visionsilter hoackinD+(varied design coModule○ Source of variant designFigure 6 Constraint flow diagram of the filter modulePhase 2: Design for temporal varietyTable 2 indicates that the critical components for realizing temporal variety arethe hierarchical graph in Fig 4, for these three components, the Carae g toHousing (part 27), Top Cover(part 1), and Carafe(part 25pies the upper level in the interaction hierarchy This arrangement means thatthe Carafe design should be addressed first, followed by that of the Housingand finally, the Cover However, the incidence and costs involved in carafedesign are quite high The strategy of Company X thus is to"over design"thiscomponent; that is, to improve the quality of the current specifications capableof handling future market requests Therefore, the Carafe is upgraded for easycleaning, pouring and dishwasher-safe in both the"current" and"future"versions

Therefore, according to the design priority the product variety shouldfocus on redesigning the Housing (part 27) To facilitate usability, the designeam tends to substitute swing-out housing for the fixed housing This changedivides the component into two new parts; namely, the Swing-out Filter Housing and the Support The Swing-out Filter Housing is further differentiatedinto either U-shaped or V-shaped The original design constraints of the Housng are laid on the two new parts, respectively(see Fig 7) Thus the shapehe Swing-out Housing must fit the Carafe; and the design of the Support must

timize variant Product Design Based on Component Interaction grafit the Base The variant design of the Housing directly influences the Top(part 1)whor convenient to use, the Top Cover is changedfrom a lift up to a fixed desiFinally, Table 6 lists the variant design driven by temporal market differentia○omponentvaried desip Fixed design constraintO Source of variant designFigure 7 Constraint flow diagram of the new designdesigned cSwing out filter basketTable 6

List of variant components for future market4 4 ResultTable 7 lists the components of the four products derived via the proposedmethodology Among these components, most of the variety occurs in chunksunks 2 and 3 remain virtually unchanged, and thus are conidered"platforms"of this product architecture Moreover, the design teamElement, Base Module, and Heating Plate Module, should be standardized to reduce the redesign effort and production co

Manufacturing the Future: Concepts, Technologies visionsNo

Product 1 Product 2 Product 3 Product 4LifiedTop cvvvvvvvvvvv-shaped fixed housinghaped fixed housingV-shapedhousing 27shaped swing out filter housing 27Water tank coveer outlet pipePacking valvevvvvvvvvvvvvvvvvvvHeating elementBascBasevvvvvvvHot plateGlass carafeThermal carafeCup bank of glass carafevvvvVvvvCup bank of thermal carafeHandle cover of glass carafehermal carafeCover of glass carafeTable 7 Components list of the product family45 Comparison of existing and proposed designsA team of engineers and managers of Company X estimated the sales volume,marketing, variable(raw material/ production prices) and fixed (engineering/jection mold) costs for the proposed designs, and compared these estimatesto those for products designed independently table 8 lists the comparison

timize variant Product Design Based on Component Interaction graThe profit is calculated using the following functionPi= Si(PRi-vCi)-FCi-MCWhere Pi, Si, PRi, VCi, FCi, MCi are the profit, sales volume, price, variable cost,fixed cost, and marketing cost of productTable 8 illustrates that the primary cost difference between the two designstrategies lies in the fixed cost The proposed designs significantly reduced thefixed cost for developing new products through sharing most componentsThe second row from the bottom shows that the profits associated with independently developing products 2 and 4 is minus 73% and 65% of currentproduct, respectively Therefore, the best decision seems to be not to developany product in Market 2

However, the proposed designs generate a totalprofit 127%nt markets and 541% in future markets higher than ifproduct 1 was the only product launched The result shows the potential saings and profit available using this methodologyNo product family designProduct family design using this methodologyof currentProduct 1 product 2 product 3 product 4 product 1 product 2 product 3 product80000080Prifuture= 14Note: VC, FC, MC are the variable, fixed and marketing costs, respectivelT

able 8 Comparison of independently developed and the proposed designtimize variant Product Design Based on Component Interaction graexplore alternative architectures through clustering high interactive components and arranging them in chunks(Pimmler Eppinger 1994, Wei 2001)Moreover, Sosa et al ,(2000)applied dSm to analyze the different typesteraction between modular and integrative systems, and Sahih Kamrani(1995) used the similarity matrix for integrating components into modulesThese studies represent crelationships in terms of similarityciprocal interaction rather than information flows However, during the embodiment design stage, variant designs of a single component can lead tomerous other components also requiring modification The hierarchicalstructure of component interactions first must be identified after which the influence of variety and subsequent design changes can be estimated To dealcomponent design constraint flows to build up feasible product architecture aawith this problem, this chapter illustrated two methodologies via identifying3 ProductBased on cor3

1 Product design rationalerved that designs are always completedthrough iteration Design iteration occurs when a new requirement is inputtedto the design task, resulting in the related components needing to be redeigned, and leading to the specifications of the other components that interactesigned components having toange thellcations todeTherefore the dm process decore and so tremendodesign efforts are required This problem becomes particularly important inanning product architectures; products must be designed to meet variousstomer needs, vet also share asible tocosts This study attempted to solve this problem by modeling component se-ential flow using ISM, interpretive structural modeling ISM is an algebratechnique for system representation and analysis that was first introduced byWarfield(1973) ISM reduces complex system interactions to a logically orented graphThis study applies and modifies ISM to establish a hierarchical component interaction structure, which can help designers to determine component commonality, variety, and design priorities

anufacturing the Future: Concepts, Technologies Visions32coprocedure of ISMPhase l: Insideratrix constructionFirst, a system is decomposed into a set of components that form a squaretrix The procedure begins with paired comparisons to identify whether a direct influence exists from component i(row)to j(column) The incidence matrix A=lail thus is defined asf a direct influence exists from component i to componenl0Fig

1(a)represents the incidence matrix of an example system containingseven components For example, the second row of the matrix indicates thatcomponent 2 directly influences components 1, 5, and 6The reachability matrix R is deducted from incidence matrix A if a boolean nmultiple product of A+l uniquely converges to R for all integers n>no, where nois an appropriate positive integer, I is a Boolean unity matrix, and is additionBoolean sense(Warfield, 1995) Matrix R represents all direct and indirectlinkages between components Figure 1(b) represents the reachability matrix Rderived from matrix A, in which an entry ri=1 if component j is reachable by i,although the path length may be one or more000(a) Original incidence matrix A(b) Reachability matrix RFigure 1 a-b Stepwise procedure of ISM

timize variant Product Design Based on Component Interaction graa technique for cluster retrieval is inserted in the ISM process to identify components that influence one another and form a loop(roberts, 1997) The reachability matrix R multiplies the transposed matrix of R, say R; thus in RRcomponents i and j mutually interact if ri ri=1 Figure 1(c) displays the outpulmatrix owhich clusters of components can be identified easily byrearranging component order Figure 1(d) reveals four clusters in the system,namely:1},12635,7and{4}0201100000000000000R·R=c3c500(c)Output matrix of R

R(d) Retrieval of clustersFigure 1 c-d, Stepwise procedure of ISMgrapFollowing cluster retrieval, the order of reachability matrix R is rearranged(asshown in Fig 1(e)), and the clustts are integrated and treated asa single entity The hierarchy graph then is obtained by identifying a setcomponents in matrix R that cannot reach or be reached by other componentsoutside the set itself, removing the set from the original matrix R, and then repeating this process for remaining matrix until a unique set of nodes that noother nodes can reach is obtained For example, in Fig 1(e), cI first is identifiedas an"exit", since it can not reach to other components; meanwhile, (c2, C6) andched by othenodes In this example, three levels of nodes were obtained (illustrated in Fi( f) The oriented links then connected the nodes from source to sink one based

on the incidence matrix Notably, the rounded rectangles in Fig 1(f)indicatethe retrieved clusters, in which the information flow forms a loop0011100000001Level 3(e)Rearranged matrix R( f)Hierarchical interactiongraph of the systemFigure 1 e-f Stepwise procedure of IsM33 Analysis procedureThe Analysis procedure comprises three main phases: market planning, QFDnd the ISM appi2phases

The first phase begins with product market planning which clarifiesthe various requirements of different markets The second phase involves theQFD analysis, during which the variant requirements are related to physicalcomponents with specific values to identify relationship degree, yielding therelative importance of each component towards the market variations Finallythe inner interactions between physical components are further examined viaarchical graph The result obtained from QFD is incorporated into the hierachical graph to identify the component to be redesigned in the influential path,deriving new products that satisfy market niches by redesigning finite compo

ent Interaction g4 Case Study for Variant Design Based on Component Interaction Graph41 Case backgrounda family of 15-liter automatic drip coffeemakers from an electronic appliances manufacturer( Company X) Ninety-fivepercent of the products of this company are original design manufact(ODM), and are mainly exported to America, Europe, and Japan Company Xms to provide product varieties to simultaneously meet the requirements ofeach segmented market, as well as to develop product architectures in masscustomization Components of the original product are listed in Table 142 Analysis procedurePhase 1: Market plThe market planning aims at two different markets(spatial variety) with twodifferent launch times(temporal variety), concurrently developing four products, as illustrated in Fig

3 The launch time of the " currentproducts isplanned for after three months, while that of"future"products is planned feafter eight monthPhase 2: Identify the exterior drivers of variationTo emphasize market differentiation the Qfd matrix lists the differences incustomer requirements rather than common requirements In the case, howmaintain coffee temperature is the key driver for spatial market differentiation,because the weather in market 2 is much colder than that of market 1 table 1illustrates thefrand 0 indicate the mapping relationships ranging from verystrong, through to strong, ordinary, weak, and none, respectively Table 1 demonstrates that the most important component for Keeping coffee temperaturethe Carafe Furthermore, the key drivers for temporal market differentiationare Ease of cleaning, Comfortable to use, and Fashionable style These re-quirements are listed in Table 2, along with their relative importance The critcal components for these requirements include the Housing, Top cover, andCarafe The QFD results are input into the product design, as described in Section 4 3

558Manufacturing the Future: Concepts, Technologies visionsrequirements related to thefic marketsThe QFD approachThe influence degree ofComponent The Ism approac DAIDdencedesign requiremeneach componentach component regardingThe degree of eacho the market-drivenHierarchical graphFigure 2 flow diagram of the aProduct 4Figure 3

Marketg of the coffee mak

timize variant Product Design Based on Component Interaction gramponentTable 1 QFD matrix of the spatially differential requirements3000000000Comfortable to300000000|170001421410s030Table 2 QFD matrix of the temporally differential requirementsPhase 3: Identify the interior hierarchical interactionsIn this approach, senior design engineers of company X perform the incidencematrix by investigating the relationships between each pair of componentsTable 3 lists the original incidence matrix The cells in the incidence matrix aremarked with "1" if the components in rows constraint the specifications of thecomponents in columns The related design constraints are documented in thent providcomponent j

For example, d(4, 5)tes that the Top Cover Base(compo-nent 5)should fit the diameter of the Spout Seat(component 4) This incidencematrix is then manipulated through the IsM procedures illustrated in Section3 2 Fig 4 shows the hierarchical graph of the design constraint flow derivedthrough ISM In this graph, the circles represent components, the oriented linesare design constraints provided by the source components, and the rounded

anufacturing the Future: Concepts, Technologies visionsrectangles indicate that a set of mutually interactive components, which are integrated as a module These modules and other components then are furthergrouped into chunks according to the frequency of their interactions Table 4lists the incidence matrix after appropriate rearrangement of the order Fourchunks are formedhe productelyhunk, C3 base chunk, and C4 carafe chunk

The precedence of the fourchunks is determined by the inter-chunk interactionsPart nan23456812l131451681902123241525272op cover sewater tank coverpacking valve 16carafe handle covfilter holder packing valve 28]er holder 30Table 3 The original incidence matrix of coffee maker components43 Design procedureThe results of the analysis illustrated in previous section are applied in theproduct design; four products were designed concurrently to satisfy requestsof different markets The design procedure is demonstrated in the following