- Email: [email protected]
A structural component-based approach for designing product family
Computers in Industry 56 (2005) 13–28 www.elsevier.com/locate/compind
A structural component-based approach for designing product family Shih-Wen Hsiao*, Elim Liu Department of Industrial Design, National Cheng Kung University, Tainan, Taiwan 70101, ROC Received 15 September 2004; accepted 28 October 2004
Abstract This study presents a novel approach for establishing an intelligent support system in order to design a product family through managing variety. Two approaches are applied successively. First, the Quality Function Deployment (QFD) approach identifies the exterior drivers of design variation. Second, the Interpretive Structural Model (ISM) technique is applied to visualize the hierarchy of component interactions within a product. This approach represents the design priority and related design constraints within a product using a structural graph, helping designers to create variant design solutions in a product family for different market requests. The methodology proposed here is illustrated via the example of the design of a family of automatic drip coffee makers. # 2004 Elsevier B.V. All rights reserved. Keywords: Product family; Interpretive structural model; Product design
1. Introduction Dominating markets with a single product is increasingly difficult and instead many industries are evolving towards mass customization, meaning the production of individually customized and highly varied products or services [1]. Some investigations have examined with the problem of product variety: * Corresponding author. Tel.: +886 6 2757575x54330; fax: +886 6 2746088. E-mail addresses: [email protected], [email protected] (S.-W. Hsiao).
Cohen [2] proposed using Master House of Quality for planning product variety. Suh [3] viewed product variety as the proper selection of design parameters that satisfy variant functional requirements. Erens [4] developed product variety under functional, technology, and physical domains. Fujita and Ishii [5] formulated the task structure of product variety design, and Martin and Ishii [6–8] proposed Design for Variety (DFV), which is a series of methodologies with quantifying indices for reducing the influence of product variety on product life-cycle cost, and thus helping design teams to develop decoupled product architectures. These studies have established a basis
0166-3615/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.compind.2004.10.002
14
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
for product variety management. However, many investigations have agreed that the key to efficiently designing and delivering multiple products is developing a product family [9–11]. The advantages of developing product family is that it enables a company to offer two or more products that are highly differentiated yet share a substantial fraction of their components. The collection of components shared by these products is called a product platform [11]. Erens [4] defined a product platform as ‘‘An architecture concept of compromising interface definitions and key-components, addressing a market and being a base for deriving different product families.’’ In his research, two examples of product platforms were addressed, one is a digital architecture for medical equipment, and the other is the operating system and the floor plan for a range of cars. In research on product families, Tseng and Jiao [12] developed the model of Product Family Architecture (PFA) in which they grouped similar products into families based on product topology and functional requirements and then established an optimal Product Family Architecture based on this grouping. Robertson and Ulrich [13] proposed a method of balancing distinctiveness with commonality through identifying the importance of various factors going into this tradeoff. Fujita et al. [14,15] utilized optimization techniques to identify the optimum architecture of a module combination across products in a family of aircraft. Moreover, Yu and Otto [16] defined Product Family Architecture based on customer needs by using the target value of product features for calculating probability distributions. Additionally, Simpson et al. [17] used the Product Platform Concept Exploration Method (PPCEM) to design a common product platform. This platform uses the market segmentation grid to help identify suitable scale factors of the platform that are ‘‘scaled’’ or ‘‘stretched’’ to satisfy various requirements. Most studies focus on optimizing family structure, rather than explicitly identifying the physical arrangement and interaction among components in the product architecture. However, during the embodiment design stage, variant designs of a single component can lead to numerous other components also requiring modification. The hierarchical structure of component interactions first must be identified, after which the influence of variety and subsequent
design changes can be estimated. To deal with this problem, this study applies a structural matrix-based method called Interpretive Structural Modeling (ISM) [18]. The purpose of the technique is to output an essential feature of a given system by applying appropriate algebraic operations to graphical expressions of the system. Some other matrix-based methods have been proposed in areas of modular product design. For example, the component-based Design Structural Matrix (DSM) method has been applied to explore alternative architectures through clustering high interactive components and arranging them in chunks [19,20]. Moreover, Sosa et al. [21] applied DSM to analyze the different types of interaction between modular and integrative systems, and Salhieh and Kamrani [22] used the similarity matrix for integrating components into modules. However, the distinguishing feature of ISM is that it represents relationships in terms of information flow rather than similarity or reciprocal interaction. This investigation attempts to facilitate designers in creating variant designs based on current product prototype, thus forming a product family to reduce redundant design effort and satisfy requirements of segmented markets. This study develops a methodology for achieving the above objective. The proposed methodology is divided into two phases; namely analysis phase and redesign phase. In analysis phase, market planning of the product family is clarified, and then the variant requirements for customers of different markets are examined and mapped into physical components through Quality Function Deployment (QFD) [23]. Meanwhile, the hierarchical interaction structure of components is established using ISM analysis, in which the design priority, and related design change resulting from component variations, is represented in graph form. The analysis results are applied in the redesign phase to help designers to concurrently create design solutions in a product family that meet different market needs without spending extra effort on redundant design circles. The rest of this paper is organized as follows. Section 2 interprets the method of ISM. Section 3 then demonstrates a case study and the analysis phases of this methodology. Next, Section 4 illustrates the redesign phase, and describes how the methodology aids designers in creating variant design and developing a product family, the results are discussed in
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Section 5. Finally, the conclusion is remarked in Section 6.
2. Method for identifying hierarchical relations among components Studies of product design have observed that designs are always completed through iteration. Design iteration occurs when a new requirement is inputted into the design task, resulting in the related components needing to be redesigned, and leading to the specifications of the other components that interact with the redesigned components having to change their specifications to fit the redesign. Therefore, the design process becomes iterative, and so tremendous design efforts are required. This problem becomes particularly important in product family development; products must be designed to meet various customer needs, yet must also share as many components as possible to maximize profits. This study attempted to solve this problem by modeling component sequential flow using ISM, interpretive structural modeling. ISM is an algebraic technique for system representation and analysis that was first introduced by Warfield [18,24– 26], and has been used in various research fields [27,28]. ISM reduces complex system interactions to a logically oriented graph. This study applies and modifies ISM to establish a hierarchical component interaction structure, which can help designers to determine component commonality, variety, and design priorities. Thus, aids designers to develop a product family. The procedures of ISM are illustrated below. (1) Incidence matrix construction. First, a system is decomposed into a set of components that form a square matrix. The procedure begins with paired comparisons to identify whether a direct influence exists from component i (row) to j (column). The incidence matrix A = [aij], thus, is defined as ( 1; if a direct influence exists from aij ¼ component i to component j 0; otherwise Fig. 1(a) represents the incidence matrix of an example system containing seven components, and displays the incidence relationships between components. For example, the second row of the
15
matrix indicates that component 2 directly influences components 1, 5, and 6. (2) Reachability matrix deduction. The reachability matrix R is deducted from incidence matrix A if a Boolean n-multiple product of A + I uniquely converges to R for all integers n > n0, where n0 is an appropriate positive integer, I a Boolean unity matrix, and + is addition in Boolean sense [25]. Matrix R represents all direct and indirect linkages between components. Relation transitivity is a basic assumption in ISM. Fig. 1(b) represents the reachability matrix R derived from matrix A, in which an entry rij = 1 if component j is reachable by i, although the path length may be one or more. (3) Cluster retrieval. A technique for cluster retrieval is inserted in the ISM process to identify components that influence one another and form a loop. Similar algorithm has been used in the graph theory [29]. The reachability matrix R multiplies the transposed matrix of R, say RT; thus in RRT, components i and j mutually interact if rijrji = 1. Fig. 1(c) displays the output matrix of RRT, in which clusters of components can be identified easily by rearranging component order. Fig. 1(d) reveals four clusters in the system, namely {1}, {2,6}, {3,5,7}, and {4}. (4) Obtained hierarchy graph. Following cluster retrieval, the order of reachability matrix R is rearranged (as shown in Fig. 1(e)), and the clustered components are integrated and treated as a single entity. The hierarchy graph then is obtained by identifying a set of components in matrix R that cannot reach or be reached by other components outside 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 no other nodes can reach is obtained. For example, in Fig. 1(e), c1 first is identified as an ‘‘exit’’, that is, the lowest level of the hierarchy, since it cannot reach to other components; meanwhile, {c2, c6} and c4 were separated as ‘‘entrances’’, because they cannot be reached by other nodes. In this example, three levels of nodes were obtained (illustrated in Fig. 1(f)). The oriented links then connected the nodes from source to sink one (for example, from c4 to c5) based on the incidence matrix. Notably, the rounded rectangles in Fig. 1(f) indicate clusters
16
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Fig. 1. Step-by-step process of ISM.
retrieved in procedure 3, in which the information flow forms a loop. In the next section, the ISM procedure is applied to visualize the hierarchy of specification flow of components within a product, and illustrated with a case study.
3. Proposed methodology: analysis phase The proposed methodology comprises three main stages: market planning, QFD and the ISM approach.
Fig. 2 presents the flow diagram for linking these stages. The first stage begins with product market planning which clarifies the various requirements of different markets. The second stage involves the QFD analysis, during which the variant requirements are related to physical components with specific values to identify relationship degree, thus yielding the relative importance of each component towards the market variations. Finally, in the third stage, the inner interactions between physical components are further examined via ISM analysis, with component design priority being represented using a hierarchical graph.
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
17
Fig. 2. Flow diagram of the proposed methodology.
The result obtained from QFD is incorporated into the hierarchical graph to identify the component to be redesigned in the influential path, deriving new products that satisfy market niches by redesigning finite components. 3.1. Object product This study employs the design of a family of 1.5-l automatic drip coffee makers to illustrate both the analysis and design phases of the proposed methodology. This case study involves a Taiwanese electronic appliances manufacturer (Company X). Ninety-five percent of the products of this company are Original Design Manufactured (ODM), and are mainly exported to America, Europe, and Japan. Based on their experiences and manufacturing technologies, Company X aims to develop product families to simultaneously meet the requirements of each segmented market, as well as to provide variety in mass customization. Fig. 3 shows an explosion view and part list of the 1.5-l automatic drip coffee maker prototype. 3.2. Identify market-driven variety 3.2.1. Market planning Market planning is performed by the product development team, which includes the marketing,
planning, and design departments of Company X. The market planning aims at two different markets (spatial variety) with two different launch times (temporal variety), concurrently developing four products, as illustrated in Fig. 4. The launch time of the ‘‘current’’ products is planned for after 3 months, while that of ‘‘future’’ products is planned for after 8 months. 3.2.2. Identify the exterior drivers of variation: the QFD analysis The QFD used in this study is modified to fit the purpose of this study, namely to identify the variant requirements of different markets, thus the QFD matrix only lists the differences in customer requirements rather than common requirements. In the present case study, because the weather in Market 2 is much colder than that of Market 1, the main concern is maintaining coffee temperature. Table 1 illustrates the mapping from requirements into components, in which the values 9, 5, 3, 1, and 0 indicate the mapping relationships ranging from very strong, through to strong, ordinary, weak, and none, respectively. Table 1 demonstrates that the most important component for keeping coffee temperature is the Carafe. Furthermore, in response to design trends, concern regarding temporal variety relates to ease of cleaning, comfortable to use, and fashionable style. These requirements are listed in the QFD matrix, along with their relative importance. Table 2 demonstrates that the critical
18
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Fig. 3. Explored view of a coffee maker prototype.
components that realize the temporal variety include the Housing, Top Cover, and Carafe. The QFD approach is helpful for identifying crucial and meaningful design changes from the perspective of
Fig. 4. Market planning of the coffee maker.
customers, and the result is input into the redesign process, as described in the following section. 3.3. Identify the interior hierarchical interactions: the ISM approach Pimmler and Eppinger [19] proposed the taxonomy interactions of components, including spatial, energy, information, and material types. Moreover, Sosa et al. [21] further identified five types of dependencies of components with the additional structural interactions. This study defines the interactions between components as design constraints, that is, the design of a component must refer to specifications of the other components related to it by using incidence matrix. This type of relationship is not always symmetric, but rather, tends to describe
Table 1 QFD matrix for mapping the spatially differential requirement into components Spatial differentiation requirement
Component
1
2
3
4
5
6
8
11
12
13
14
15
16
18
19
20
21
22
23
24
25
26
27
28
30
Keep coffee 0 temperature
0
0
0
0
0
0
0
0
0
0
0
0
3
0
1
3
0
0
0
9
3
0
0
0
Table 2 QFD matrix for mapping the temporal differential requirements into components Temporal Component differentiation requirement Top Top Spout Spout Top Water Water Base Silicone Water Pipe Base Packing Heating Switch Hot Hot Cup Carafe Carafe Carafe Carafe Housing Filter Filter Relative Cover Cover Seat Cover Tank Tank ring outlet connection Cover Valve Element plate plate Bank Handle Handle Cover Holder Holder weight Base Base Cover pipe seat ring Cover Packing Valve 1 Ease of cleaning Comfortable to use Fashionable style Total
2
3
4
11
12
13
14
15
16
18
19
20
21
22
23
24
25
26
27
28
30
1
3
0
0
5 3
6 1
8 3
0
0
0
0
0
0
0
0
0
0
0
1
0
5
1
1
0
5
3
9
0
1
1
1
3
0
0
0
0
0
0
0
0
3
0
0
0
1
3
3
1
9
0
3
5
3
0
0
0
0
1
3
3
0
0
0
0
0
0
1
0
0
0
3
3
9
1
5
0
0
2
54
9
5
5
14
20
15
6
0
0
0
0
0
0
17
0
0
0
14
21
48
10
58
0
30
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Top Top Spout Spout Top Water Water Base Silicone Water Pipe Base Packing Heating Switch Hot Hot Cup Carafe Carafe Carafe Carafe Housing Filter Filter Cover Cover Seat Cover Tank Tank ring Outlet connection Cover Valve Element plate plate Bank Handle handle Cover Holder Holder Base Base Cover pipe seat ring Cover Packing Valve
19
20
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 3 The original incidence matrix of coffee maker components
component interaction from the perspective of design information supply-receipt. In the ISM approach, senior design engineers of Company X perform the incidence matrix by investigating the relationships between each pair of components. Table 3 lists the original incidence matrix. The cells in the incidence matrix are marked with ‘‘1’’ if the components in rows constraint the specifications of the components in columns. The related design constraints are documented in the form d(i, j), where i denotes the source component providing a constraint to component j. For example, d(4, 5) indicates that the Top Cover Base (component 5) should fit the diameter of the Spout Seat
(component 4). The identified relations represent design constraints and incidence between components that cope with the design knowledge of the specific product. The rightmost column in Table 3 lists the D value (sum of columns), indicating the degree to which each component influences others. The bottom row in the table shows the R value (sum of rows), indicating the degree to which each component is influenced by the others. This incidence matrix is then manipulated through the ISM procedures illustrated in Section 2. Fig. 5 shows the hierarchical graph of the design constraint flow derived through ISM. In this graph, the circles represent components, the oriented lines are design constraints provided by the source components,
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
21
Fig. 6. Plotted diagram of component interaction.
(D + R), this graph is an overall indicator of how interactive/dominant a component is. For example, the high (D R) value indicates that changes to the component have relatively high propagation strength, and a high (D + R) component is an interface component; changes to which affect or refer to numerous components. Fig. 5. Hierarchical graph of component interaction.
and the rounded rectangles indicate that a set of mutually interactive components, which are integrated as a module. These modules and other components then are further grouped into chunks according to the frequency of their interactions. Table 4 lists the incidence matrix after appropriate rearrangement of the order, and four chunks are formed in the product, namely C1 housing chunk, C2 water tank chunk, C3 base chunk, and C4 carafe chunk. The precedence of the four chunks is determined by the inter-chunk interactions. Furthermore, applying the values of D and R shown in Table 3, the last two columns of Table 4 list the values of (D + R) and (D R), respectively. The (D + R) value indicates the sum of interactions of a component, including the ‘flow out’ and ‘flow in’ interactions. The (D R) indicates the difference between the influencing and influenced interactions of a component, and a higher value indicates that the component is dominant rather than subordinate. For example, Table 4 reveals that the highest (D + R) value is 10 for component 11, namely the Base. The highest (D R) value is 5 for component 25, namely the Carafe. Fig. 6 shows the (D R) plotted against
4. Applying the analytical result for product family development: redesign phase In the redesign phase, the results of the analysis of QFD and ISM illustrated in previous section are applied in product design and product family development. In this stage, four products were designed concurrently to satisfy requests of different markets. The design procedures are demonstrated in the following paragraphs. 4.1. Design for spatial variety The QFD approach listed in Table 1 indicates that Carafe (part no. 25) redesign is essential for addressing concerns regarding market variety. In Market 2, to maintain coffee temperature, the Carafe must use heat insulation material; moreover, the wall should be thickened, the shape of Carafe slenderized and the top narrowed to reduce heat loss. These changes of carafe specification also affect the design of other components. The incidence relation diagram of the Carafe is shown in Fig. 7 (extracted from Fig. 5). In this figure, the Carafe is the source component for the variant design (represented as a double-circle),
22
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 4 Incidence matrix after appropriate rearranging the order
Note: Grayed cells indicate the inter-chunk interactions.
exporting design constraints to the sink nodes. The design constraints are listed below: d(25, 21): The Heating Plate Module should fit the diameter of Carafe Base (fixed). d(25, 22): The Cup Bank should fit the diameter of Carafe body (changed). d(25, 24): The Carafe Handle should fit the arc and weight of Carafe body (changed). d(25, 26): The Carafe Cover should fit the diameter of Carafe rim, and the requested thermal condition (changed). d(25, 27): The Filter Module should fit the Carafe height (changed).
Fig. 7. Incidence diagram of Carafe (component 25).
The constraint of d(25, 21) is fixed (represented as dotted line in Fig. 7), and thus parts (20, 21) are
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
23
Under these constraints, the design of Filter Module (parts 27, 5, and 30) is changed from V- to Ushaped to fit the new Carafe design. Furthermore, constraint from the Filter Module is: d(5, 1): The Top Cover should fit the Basket Holder rim diameter. Since the specification of the Basket Holder rim is fixed, component 1 and 2 need not change their design. Consequently, Table 5 lists the variant component design driven by spatial market variety. 4.2. Design for temporal variety
Fig. 8. Constraint flow diagram of the Filter Module.
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. 7) owing to the variant carafe specification, resulting in the design of the Filter Module and Carafe Outfit Module having to be changed to match the altered conditions. In the Carafe Outfit Module, the components are redesigned to fit the new Carafe. However, the design change of the Filter Module must refer not only to the Carafe, but also to other components that provide constraints on the Filter Module, as shown in Fig. 8. Thus, in redesigning the Filter Module, the constraint from the Carafe becomes the source of variant design (represented as solid line in Fig. 8), while the others are fixed constraints (represented as dotted lines in Fig. 8) listed below: d(11, 27): The Housing should fit the Base. d(28, 30): The Filter Holder should fit the Filter Holder Packing Valve diameter. d(4, 5): The Top Cover Base should fit the Spout Seat diameter. d(3, 30): The Filter Holder should fit the Spout shape.
Table 2 indicates that the critical components for realizing temporal variety are the Housing (part 27), Top Cover (part 1), and Carafe (part 25). According to the hierarchical graph in Fig. 5, for these three components, the Carafe occupies the upper level in the interaction hierarchy. This arrangement means that the Carafe redesign should be addressed first, followed by that of the Housing and finally, the Cover. However, the incidence and costs involved in carafe redesign are quite high. The strategy of Company X thus is to ‘‘over-design’’ this component; that is, to improve the quality of the current specifications capable of handling future market requests. Therefore, the Carafe is upgraded for easy cleaning, pouring and dishwasher-safe, and both the ‘‘current’’ and ‘‘future’’ versions of the component are left unchanged. Therefore, according to the design priority, the product variety should focus on redesigning the Housing (part 27). To facilitate usability, the design team tends to substitute Swing-out Housing for the fixed housing. This change divides the component into two new parts; namely, the Swing-out Filter Housing and the Support. The Swing-out Filter Housing is further differentiated into either U- or V-shaped. The original design constraints of the Housing are laid on the two new parts, respectively (see Fig. 9). Thus, the shape of the Swing-out Housing thus must fit the Carafe; and the design of the Support must fit the Base. The variant design of the Housing directly influences the Top Cover (part 1); meanwhile, for convenient to use, the Top Cover is changed from a lift-up to a fixed design.
24
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 5 List of variant design driven by spatial variety Number
Component
Market 1
25 25* 22 22* 23 23* 24 24* 26 26* 27 27* 5 5* 30 30*
Glass carafe Thermal carafe Cup Bank of glass carafe Cup Bank of thermal carafe Handle Cover of glass carafe Handle Cover of thermal carafe Handle of glass carafe Handle of thermal carafe Cover of glass carafe Cover of thermal carafe V-shaped housing U-shaped housing V-shaped Cover Base U-shaped Cover Base V-shaped filter basket U-shaped filter basket
V
Market 2 V
V V V V V V V V V V V V V V
Note: *New design of the component.
Finally, Table 6 lists the variant design driven by temporal variety. 4.3. Developing a product family Four family products corresponding to different competitive conditions are derived based on the
variant designs developed above, the schematic diagram is illustrated in Fig. 10. Table 7 lists the components of the four products. Among these components, most of the variety occurs in chunks 1 and 4, while chunks 2 and 3 remain virtually unchanged, and thus are considered ‘‘platforms’’ of this product family. Moreover, the design team further suggested that components of the upper levels of chunk 3, including Water pipe Module, Heating Element, Base Module, and Heating Plate Module, should be standardized to reduce the redesign effort and production cost. A team of engineers and top managers of Company X estimated the sales volume, marketing, variable (raw material/production prices), and fixed (engineering/injection mold) costs for each product of using the product family designs, and compared these estimates
Table 6 List of variant design driven by temporal variety Number
Component
Current
27 27*1 27*2 1 1*
Housing Swing-out Filter Basket Support Lift-up Top Cover Fixed Top Cover
V
*
Fig. 9. Constraint flow diagram of the new design.
Future V V
V V *1
*2
Note: New design of the component. 27 and 27 are the two new designs of component 27.
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
25
Fig. 10. Schematic diagram of the final result.
Table 7 Components list of the product family Chunk
C1
Component
Number
Product 1
Product 2
Lift-up Top Cover Fixed Top Cover Top Cover set Spout Spout Seat Filter Holder Packing Valve V-shaped Cover Base U-shaped Cover Base V-shaped Filter Holder U-shaped Filter Holder V-shaped fixed housing U-shaped fixed housing V-shaped Swing-out Filter Housing U-shaped Swing-out Filter Housing
1 1* 2 3 4 28 5 5* 30 30* 27 27* 27*11 27*12
V
V
V V V V V
V V V V
Support Water Tank Cover
C2
Water Tank Silicone ring Water outlet pipe Pipe connection seat Packing Valve Heating Element Base Base Cover Switch Hot plate ring
C3
Hot plate Glass carafe Thermal carafe
Product 3
Product 4
V V V V V V
V V V V V
V V
V V
V
V
V V V V
*2
6
V
V
V V
V V
8 12 13 14 16 18 11 15 19 20
V V V V V V V V V V
V V V V V V V V V V
V V V V V V V V V V
V V V V V V V V V V
21 25 25*
V V
V
V V
V
27
V
V
26
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 7 (Continued ) Chunk
Component
C4 *11
Note: 27
Number
Cup Bank of glass carafe Cup Bank of thermal carafe Handle Cover of glass carafe Handle Cover of thermal carafe Handle of glass carafe Handle of thermal carafe Cover of glass carafe
22 22* 23 23* 24 24* 26
Cover of thermal carafe
26*
*12
and 27
Product 1
Product 2
Product 3
V
Product 4
V V
V
V
V V
V
V
V V
V
V
V V
V
are new designs derived from component 27*1.
Table 8 Comparing costs and profit between independently developed and family product design % of current product
Sales volume Price VC FC MC Profit Total profit
No product family design
Product family design using this methodology
Product 1
Product 2
Product 3
Product 4
Product 1
Product 2
Product 3
Product 4
100 100 100 100 100 100
80 120 145 100 200 73
100 115 105 100 100 208
80 130 145 100 100 65
100 100 100 100 100 100
80 120 140 40 200 27
100 115 100 10 100 334
80 130 140 10 100 207
Current = 27
Future = 143
Current = 127
Future = 541
Note: VC, FC, and MC are the variable, fixed, and marketing costs, respectively.
to those for products designed independently. Table 8 lists the comparison. The profit is calculated using the following function: Pi ¼ Si ðPRi VCi Þ FCi MCi
only product launched. The result shows the potential savings and profit available using this methodology.
(1)
where Pi, Si, PRi, VCi, FCi, MCi are the profit, sales volume, price, variable cost, fixed cost, and marketing cost of product i, respectively. Table 8 illustrates that the primary cost difference between the two design strategies lies in the fixed cost. The product family designs significantly reduced the fixed cost for developing new products through sharing most components. The second row from the bottom shows that the profits associated with independently developing products 2 and 4 is minus 73% and 65% of current product, respectively. Therefore, the best decision seems to be not to develop any product in Market 2. However, product designs using this methodology generate a total profit 127% in current markets and 541% in future markets higher than if product 1 was the
5. Discussion Based on this investigation and application involving practical product design, for product family design through variety management, the hierarchical graph could support the following design strategies: (1) Design strategies for the dominant components: (a) The drivers of design variation of these components are customer requirements. Since the incidence and effort for the design variation are relative huge, this variation must be obvious to the customers, and customer requirements are likely to be considered as completely as possible. To achieve a stable Product Family Architecture,
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
‘‘over-design’’ of the components might be unavoidable in response to upgrade requests. (b) If the components are remained unchanged, they should be considered to be standardized or fixed specification, and become core platform of the product family. (2) The designs of the sink components are more likely to change to comply with both the altered design constraints and the requests of customer requirements. However, the cost and incidence of altering these components is relatively low. Furthermore, the redesign should incorporate the constraints provided by the source components. (3) Components that with their specification flow forms an interaction loop are likely to be further modularized or integrated.
6. Conclusions This study structurally illustrated the product family formation. The system employs two techniques: the ISM and QFD. Applications of these techniques to product architecture provide the hierarchical graph of interactions. Furthermore, the proposed methodology provides the following advantages for developing product family: (1) The methodology clarifies the specification flow between components rather than merely symmetric relationship ‘‘similarity’’ or ‘‘correlation’’. Thus, the necessary information is provided for determining not only clustering but also precedence among components. (2) The incidence matrix with the documented design constraints provides a computable way for design knowledge representation. (3) The hierarchical graphical diagram provides designers with a user-friendly display for clarifying the influence of each component variation. (4) The hierarchical structure along with the QFD analysis helps product family developers to identify whether the components should be standardized, altered, or modularized. (5) The integration of spatial and temporal market requirements is beneficial for establishing product family without redundant development costs and increasing product development lead time.
27
(6) The proposed approach can be integrated into a design information system (DIS) in the future for enhancing collaborative design and development.
References [1] B.J. Pine, Mass Customization: The New Frontier in Business Competition, Harvard Business School Press, Boston, 1993. [2] L. Cohen, Quality Function Deployment: How to Make QFD Work for You, Addison-Weesley, Reading, MA, 1995. [3] N.P. Suh, The Principles of Design, Oxford University Press, NY, 1990. [4] F.J. Erens, The synthesis of variety: developing product families, Dissertation Eindhoven University of Technology, 1996. [5] K. Fujita, K. Ishii, Task structuring toward computational approaches to product variety design, in: Proceedings of the ASME Design Engineering Technical Conferences, 1997, DETC97/DAC-3766. [6] M.V. Martin, K. Ishii, Design for variety: a methodology for understanding the costs of product proliferation, in: Proceedings of the ASME Design Engineering Technical Conference, 1996, 96-DETC/DTM-1610. [7] M.V. Martin, K. Ishii, Design for variety: development of complexity indices and design chart, in: Proceedings of the ASME Design Engineering Technical Conference, 1997, DETC97/DFM-4359. [8] M.V. Martin, K. Ishii, Design for variety: developing standardized and modularized product platform architectures, Research in Engineering Design 13 (2002) 213–235. [9] M.H. Meyer, J.M. Utterback, The product family and the dynamics of core capability, Sloan Management Review 34 (1993) 29–47. [10] M.S. Sawhney, Leveraged high-variety strategies: from portfolio thinking to platform thinking, Journal of the Academy of Marketing Science 26 (1998) 54–61. [11] K.T. Ulrich, S.D. Eppinger, Product Design and Development, second ed. McGraw-Hill, New York, 2000. [12] M.M. Tseng, J. Jiao, Design for mass customization by developing product family architecture, in: Proceedings of the ASME Design Engineering Technical Conference, 1998, DETC98/DTM-5717. [13] D. Robertson, K.T. Ulrich, Planning for product platforms, Sloan Management Review 39 (1998) 19–31. [14] K. Fujita, S. Akagi, T. Yoneda, M. Ishikawa, Simultaneous optimization of product family sharing system structure and configuration, in: Proceedings of the ASME Design Engineering Technical Conferences, 1998, DETC98/DFM-5722. [15] K. Fujita, H. Sakaguchi, S. Akagi, Product variety deployment and its optimization under modular architecture and module commonalization, in: Proceedings of the ASME Design Engineering Technical Conferences, 1999, DETC99/DFM-8923. [16] J.S. Yu, K.N. Otto, Product architecture definition based on customer demands, in: Proceedings of the ASME Int Design Engineering Technical Conferences, 1998, DETC98/DTM5679.
28
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
[17] T.W. Simpson, J.R.A. Maier, F. Mistree, A Product Platform Concept Exploration Method for Product Family Design, in: Proceedings of the ASME Design Engineering Technical Conference, 1999, DETC99/DTM-8761. [18] J.N. Warfield, Societal Systems: Planning, Policy and Complexity, Wiley, New York, 1976. [19] Y.U. Pimmler, S.D. Eppinger, Integration analysis of product decomposition, Proceedings of the ASME Design Theory and Methodology Conference 68 (1994) 343–351. [20] M.J.V. Wei, J.L. Greer, M.I. Campbell, R.B. Atone, K.L. Wood, Interfaces and product architecture, in: Proceedings of the ASME International Design Engineering Technical Conferences, 2001, DETC01/DTM-21689. [21] M.E. Sosa, S.D. Eppinger, C.M. Rowles, Designing modular and integrative systems, in: Proceedings of the ASME Design Engineering Technical Conference, 2000, DETC2000/DTM14571. [22] S.M. Salhieh, A.K. Kamrani, Macro level product development using design for modularity, Robotics and Computer Integrated Manufacturing 15 (1999) 319–329. [23] J. Hauser, D. Clausing, The house of quality, Harvard Business Review 66 (1988) 63–73. [24] J.N. Warfield, On arranging elements of a hierarchy in graphic form, IEEE-SMC 3 (1973) 121–132. [25] J.N. Warfield, A Science of Generic Design: Managing Complexity Through System Design, vol. 1, Intersystems, Salinas, CA, 1990. [26] J.N. Warfield, A Science of Generic Design: Managing Complexity Through System Design, vol. 2, Intersystems, Salinas, CA, 1990. [27] T. Kitamura, A computer-aided approach to the structural analysis and modification of a large circulatory system model, IEEE Transaction on Biomedical Engineering 46 (1999) 485– 493. [28] H.D. Sharma, A.D. Gupta, Sushil, The objectives of waste management in India: a futures inquiry, Technological Forecasting and Social Change 48 (1995) 285–309.
[29] F.S. Roberts, Discrete Mathematical Models, Prentice-Hall Englewood Cliffs, NJ, 1976. Shih Wen Hsiao is a professor in the Department of Industrial Design at the National Cheng Kung University, Tainan, Taiwan, Republic of China. He was also appointed to the Chairman of this Department in the period of 1998–2001. He received his PhD in mechanical engineering from National Cheng Kung University in 1990. Before joining the University in 1991, he spent about 13 years in the China Steel Corporation. His major research interests include fuzzy set theory, concurrent engineering, computer-aided design, neural network, genetic algorithm, gray theory, color planning for product design, heat transfer analysis, and application of reverse engineering in product design. He has published a lot of papers in the international journals in the above-mentioned areas. He was invited as a Keynote Speaker for the Conference on Grey System Theory and Its Applications, which was held in 2001. He got an Outstanding Research Award from Engineering College in 2002 and got a Distinguished Professor Award at National Cheng Kung University in 2003. Dr. Hsiao has been profiled in Who’s Who in the World, Who’s Who in Design, 500 Great Minds of the Early 21st Century, and 500 Distinguished Professors and Scholar of the BWW Society. Elim Liu is a PhD candidate in the Department of Industrial Design, National Cheng Kung University, Taiwan, Republic of China. She received her Bachelor degree and Master degree in Industrial Design from National Cheng Kung University in 1990 and 1993. Her research interests include product family development, and artificial intelligent based design inference system.
A structural component-based approach for designing product family Shih-Wen Hsiao*, Elim Liu Department of Industrial Design, National Cheng Kung University, Tainan, Taiwan 70101, ROC Received 15 September 2004; accepted 28 October 2004
Abstract This study presents a novel approach for establishing an intelligent support system in order to design a product family through managing variety. Two approaches are applied successively. First, the Quality Function Deployment (QFD) approach identifies the exterior drivers of design variation. Second, the Interpretive Structural Model (ISM) technique is applied to visualize the hierarchy of component interactions within a product. This approach represents the design priority and related design constraints within a product using a structural graph, helping designers to create variant design solutions in a product family for different market requests. The methodology proposed here is illustrated via the example of the design of a family of automatic drip coffee makers. # 2004 Elsevier B.V. All rights reserved. Keywords: Product family; Interpretive structural model; Product design
1. Introduction Dominating markets with a single product is increasingly difficult and instead many industries are evolving towards mass customization, meaning the production of individually customized and highly varied products or services [1]. Some investigations have examined with the problem of product variety: * Corresponding author. Tel.: +886 6 2757575x54330; fax: +886 6 2746088. E-mail addresses: [email protected], [email protected] (S.-W. Hsiao).
Cohen [2] proposed using Master House of Quality for planning product variety. Suh [3] viewed product variety as the proper selection of design parameters that satisfy variant functional requirements. Erens [4] developed product variety under functional, technology, and physical domains. Fujita and Ishii [5] formulated the task structure of product variety design, and Martin and Ishii [6–8] proposed Design for Variety (DFV), which is a series of methodologies with quantifying indices for reducing the influence of product variety on product life-cycle cost, and thus helping design teams to develop decoupled product architectures. These studies have established a basis
0166-3615/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.compind.2004.10.002
14
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
for product variety management. However, many investigations have agreed that the key to efficiently designing and delivering multiple products is developing a product family [9–11]. The advantages of developing product family is that it enables a company to offer two or more products that are highly differentiated yet share a substantial fraction of their components. The collection of components shared by these products is called a product platform [11]. Erens [4] defined a product platform as ‘‘An architecture concept of compromising interface definitions and key-components, addressing a market and being a base for deriving different product families.’’ In his research, two examples of product platforms were addressed, one is a digital architecture for medical equipment, and the other is the operating system and the floor plan for a range of cars. In research on product families, Tseng and Jiao [12] developed the model of Product Family Architecture (PFA) in which they grouped similar products into families based on product topology and functional requirements and then established an optimal Product Family Architecture based on this grouping. Robertson and Ulrich [13] proposed a method of balancing distinctiveness with commonality through identifying the importance of various factors going into this tradeoff. Fujita et al. [14,15] utilized optimization techniques to identify the optimum architecture of a module combination across products in a family of aircraft. Moreover, Yu and Otto [16] defined Product Family Architecture based on customer needs by using the target value of product features for calculating probability distributions. Additionally, Simpson et al. [17] used the Product Platform Concept Exploration Method (PPCEM) to design a common product platform. This platform uses the market segmentation grid to help identify suitable scale factors of the platform that are ‘‘scaled’’ or ‘‘stretched’’ to satisfy various requirements. Most studies focus on optimizing family structure, rather than explicitly identifying the physical arrangement and interaction among components in the product architecture. However, during the embodiment design stage, variant designs of a single component can lead to numerous other components also requiring modification. The hierarchical structure of component interactions first must be identified, after which the influence of variety and subsequent
design changes can be estimated. To deal with this problem, this study applies a structural matrix-based method called Interpretive Structural Modeling (ISM) [18]. The purpose of the technique is to output an essential feature of a given system by applying appropriate algebraic operations to graphical expressions of the system. Some other matrix-based methods have been proposed in areas of modular product design. For example, the component-based Design Structural Matrix (DSM) method has been applied to explore alternative architectures through clustering high interactive components and arranging them in chunks [19,20]. Moreover, Sosa et al. [21] applied DSM to analyze the different types of interaction between modular and integrative systems, and Salhieh and Kamrani [22] used the similarity matrix for integrating components into modules. However, the distinguishing feature of ISM is that it represents relationships in terms of information flow rather than similarity or reciprocal interaction. This investigation attempts to facilitate designers in creating variant designs based on current product prototype, thus forming a product family to reduce redundant design effort and satisfy requirements of segmented markets. This study develops a methodology for achieving the above objective. The proposed methodology is divided into two phases; namely analysis phase and redesign phase. In analysis phase, market planning of the product family is clarified, and then the variant requirements for customers of different markets are examined and mapped into physical components through Quality Function Deployment (QFD) [23]. Meanwhile, the hierarchical interaction structure of components is established using ISM analysis, in which the design priority, and related design change resulting from component variations, is represented in graph form. The analysis results are applied in the redesign phase to help designers to concurrently create design solutions in a product family that meet different market needs without spending extra effort on redundant design circles. The rest of this paper is organized as follows. Section 2 interprets the method of ISM. Section 3 then demonstrates a case study and the analysis phases of this methodology. Next, Section 4 illustrates the redesign phase, and describes how the methodology aids designers in creating variant design and developing a product family, the results are discussed in
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Section 5. Finally, the conclusion is remarked in Section 6.
2. Method for identifying hierarchical relations among components Studies of product design have observed that designs are always completed through iteration. Design iteration occurs when a new requirement is inputted into the design task, resulting in the related components needing to be redesigned, and leading to the specifications of the other components that interact with the redesigned components having to change their specifications to fit the redesign. Therefore, the design process becomes iterative, and so tremendous design efforts are required. This problem becomes particularly important in product family development; products must be designed to meet various customer needs, yet must also share as many components as possible to maximize profits. This study attempted to solve this problem by modeling component sequential flow using ISM, interpretive structural modeling. ISM is an algebraic technique for system representation and analysis that was first introduced by Warfield [18,24– 26], and has been used in various research fields [27,28]. ISM reduces complex system interactions to a logically oriented graph. This study applies and modifies ISM to establish a hierarchical component interaction structure, which can help designers to determine component commonality, variety, and design priorities. Thus, aids designers to develop a product family. The procedures of ISM are illustrated below. (1) Incidence matrix construction. First, a system is decomposed into a set of components that form a square matrix. The procedure begins with paired comparisons to identify whether a direct influence exists from component i (row) to j (column). The incidence matrix A = [aij], thus, is defined as ( 1; if a direct influence exists from aij ¼ component i to component j 0; otherwise Fig. 1(a) represents the incidence matrix of an example system containing seven components, and displays the incidence relationships between components. For example, the second row of the
15
matrix indicates that component 2 directly influences components 1, 5, and 6. (2) Reachability matrix deduction. The reachability matrix R is deducted from incidence matrix A if a Boolean n-multiple product of A + I uniquely converges to R for all integers n > n0, where n0 is an appropriate positive integer, I a Boolean unity matrix, and + is addition in Boolean sense [25]. Matrix R represents all direct and indirect linkages between components. Relation transitivity is a basic assumption in ISM. Fig. 1(b) represents the reachability matrix R derived from matrix A, in which an entry rij = 1 if component j is reachable by i, although the path length may be one or more. (3) Cluster retrieval. A technique for cluster retrieval is inserted in the ISM process to identify components that influence one another and form a loop. Similar algorithm has been used in the graph theory [29]. The reachability matrix R multiplies the transposed matrix of R, say RT; thus in RRT, components i and j mutually interact if rijrji = 1. Fig. 1(c) displays the output matrix of RRT, in which clusters of components can be identified easily by rearranging component order. Fig. 1(d) reveals four clusters in the system, namely {1}, {2,6}, {3,5,7}, and {4}. (4) Obtained hierarchy graph. Following cluster retrieval, the order of reachability matrix R is rearranged (as shown in Fig. 1(e)), and the clustered components are integrated and treated as a single entity. The hierarchy graph then is obtained by identifying a set of components in matrix R that cannot reach or be reached by other components outside 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 no other nodes can reach is obtained. For example, in Fig. 1(e), c1 first is identified as an ‘‘exit’’, that is, the lowest level of the hierarchy, since it cannot reach to other components; meanwhile, {c2, c6} and c4 were separated as ‘‘entrances’’, because they cannot be reached by other nodes. In this example, three levels of nodes were obtained (illustrated in Fig. 1(f)). The oriented links then connected the nodes from source to sink one (for example, from c4 to c5) based on the incidence matrix. Notably, the rounded rectangles in Fig. 1(f) indicate clusters
16
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Fig. 1. Step-by-step process of ISM.
retrieved in procedure 3, in which the information flow forms a loop. In the next section, the ISM procedure is applied to visualize the hierarchy of specification flow of components within a product, and illustrated with a case study.
3. Proposed methodology: analysis phase The proposed methodology comprises three main stages: market planning, QFD and the ISM approach.
Fig. 2 presents the flow diagram for linking these stages. The first stage begins with product market planning which clarifies the various requirements of different markets. The second stage involves the QFD analysis, during which the variant requirements are related to physical components with specific values to identify relationship degree, thus yielding the relative importance of each component towards the market variations. Finally, in the third stage, the inner interactions between physical components are further examined via ISM analysis, with component design priority being represented using a hierarchical graph.
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
17
Fig. 2. Flow diagram of the proposed methodology.
The result obtained from QFD is incorporated into the hierarchical graph to identify the component to be redesigned in the influential path, deriving new products that satisfy market niches by redesigning finite components. 3.1. Object product This study employs the design of a family of 1.5-l automatic drip coffee makers to illustrate both the analysis and design phases of the proposed methodology. This case study involves a Taiwanese electronic appliances manufacturer (Company X). Ninety-five percent of the products of this company are Original Design Manufactured (ODM), and are mainly exported to America, Europe, and Japan. Based on their experiences and manufacturing technologies, Company X aims to develop product families to simultaneously meet the requirements of each segmented market, as well as to provide variety in mass customization. Fig. 3 shows an explosion view and part list of the 1.5-l automatic drip coffee maker prototype. 3.2. Identify market-driven variety 3.2.1. Market planning Market planning is performed by the product development team, which includes the marketing,
planning, and design departments of Company X. The market planning aims at two different markets (spatial variety) with two different launch times (temporal variety), concurrently developing four products, as illustrated in Fig. 4. The launch time of the ‘‘current’’ products is planned for after 3 months, while that of ‘‘future’’ products is planned for after 8 months. 3.2.2. Identify the exterior drivers of variation: the QFD analysis The QFD used in this study is modified to fit the purpose of this study, namely to identify the variant requirements of different markets, thus the QFD matrix only lists the differences in customer requirements rather than common requirements. In the present case study, because the weather in Market 2 is much colder than that of Market 1, the main concern is maintaining coffee temperature. Table 1 illustrates the mapping from requirements into components, in which the values 9, 5, 3, 1, and 0 indicate the mapping relationships ranging from very strong, through to strong, ordinary, weak, and none, respectively. Table 1 demonstrates that the most important component for keeping coffee temperature is the Carafe. Furthermore, in response to design trends, concern regarding temporal variety relates to ease of cleaning, comfortable to use, and fashionable style. These requirements are listed in the QFD matrix, along with their relative importance. Table 2 demonstrates that the critical
18
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Fig. 3. Explored view of a coffee maker prototype.
components that realize the temporal variety include the Housing, Top Cover, and Carafe. The QFD approach is helpful for identifying crucial and meaningful design changes from the perspective of
Fig. 4. Market planning of the coffee maker.
customers, and the result is input into the redesign process, as described in the following section. 3.3. Identify the interior hierarchical interactions: the ISM approach Pimmler and Eppinger [19] proposed the taxonomy interactions of components, including spatial, energy, information, and material types. Moreover, Sosa et al. [21] further identified five types of dependencies of components with the additional structural interactions. This study defines the interactions between components as design constraints, that is, the design of a component must refer to specifications of the other components related to it by using incidence matrix. This type of relationship is not always symmetric, but rather, tends to describe
Table 1 QFD matrix for mapping the spatially differential requirement into components Spatial differentiation requirement
Component
1
2
3
4
5
6
8
11
12
13
14
15
16
18
19
20
21
22
23
24
25
26
27
28
30
Keep coffee 0 temperature
0
0
0
0
0
0
0
0
0
0
0
0
3
0
1
3
0
0
0
9
3
0
0
0
Table 2 QFD matrix for mapping the temporal differential requirements into components Temporal Component differentiation requirement Top Top Spout Spout Top Water Water Base Silicone Water Pipe Base Packing Heating Switch Hot Hot Cup Carafe Carafe Carafe Carafe Housing Filter Filter Relative Cover Cover Seat Cover Tank Tank ring outlet connection Cover Valve Element plate plate Bank Handle Handle Cover Holder Holder weight Base Base Cover pipe seat ring Cover Packing Valve 1 Ease of cleaning Comfortable to use Fashionable style Total
2
3
4
11
12
13
14
15
16
18
19
20
21
22
23
24
25
26
27
28
30
1
3
0
0
5 3
6 1
8 3
0
0
0
0
0
0
0
0
0
0
0
1
0
5
1
1
0
5
3
9
0
1
1
1
3
0
0
0
0
0
0
0
0
3
0
0
0
1
3
3
1
9
0
3
5
3
0
0
0
0
1
3
3
0
0
0
0
0
0
1
0
0
0
3
3
9
1
5
0
0
2
54
9
5
5
14
20
15
6
0
0
0
0
0
0
17
0
0
0
14
21
48
10
58
0
30
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Top Top Spout Spout Top Water Water Base Silicone Water Pipe Base Packing Heating Switch Hot Hot Cup Carafe Carafe Carafe Carafe Housing Filter Filter Cover Cover Seat Cover Tank Tank ring Outlet connection Cover Valve Element plate plate Bank Handle handle Cover Holder Holder Base Base Cover pipe seat ring Cover Packing Valve
19
20
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 3 The original incidence matrix of coffee maker components
component interaction from the perspective of design information supply-receipt. In the ISM approach, senior design engineers of Company X perform the incidence matrix by investigating the relationships between each pair of components. Table 3 lists the original incidence matrix. The cells in the incidence matrix are marked with ‘‘1’’ if the components in rows constraint the specifications of the components in columns. The related design constraints are documented in the form d(i, j), where i denotes the source component providing a constraint to component j. For example, d(4, 5) indicates that the Top Cover Base (component 5) should fit the diameter of the Spout Seat
(component 4). The identified relations represent design constraints and incidence between components that cope with the design knowledge of the specific product. The rightmost column in Table 3 lists the D value (sum of columns), indicating the degree to which each component influences others. The bottom row in the table shows the R value (sum of rows), indicating the degree to which each component is influenced by the others. This incidence matrix is then manipulated through the ISM procedures illustrated in Section 2. Fig. 5 shows the hierarchical graph of the design constraint flow derived through ISM. In this graph, the circles represent components, the oriented lines are design constraints provided by the source components,
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
21
Fig. 6. Plotted diagram of component interaction.
(D + R), this graph is an overall indicator of how interactive/dominant a component is. For example, the high (D R) value indicates that changes to the component have relatively high propagation strength, and a high (D + R) component is an interface component; changes to which affect or refer to numerous components. Fig. 5. Hierarchical graph of component interaction.
and the rounded rectangles indicate that a set of mutually interactive components, which are integrated as a module. These modules and other components then are further grouped into chunks according to the frequency of their interactions. Table 4 lists the incidence matrix after appropriate rearrangement of the order, and four chunks are formed in the product, namely C1 housing chunk, C2 water tank chunk, C3 base chunk, and C4 carafe chunk. The precedence of the four chunks is determined by the inter-chunk interactions. Furthermore, applying the values of D and R shown in Table 3, the last two columns of Table 4 list the values of (D + R) and (D R), respectively. The (D + R) value indicates the sum of interactions of a component, including the ‘flow out’ and ‘flow in’ interactions. The (D R) indicates the difference between the influencing and influenced interactions of a component, and a higher value indicates that the component is dominant rather than subordinate. For example, Table 4 reveals that the highest (D + R) value is 10 for component 11, namely the Base. The highest (D R) value is 5 for component 25, namely the Carafe. Fig. 6 shows the (D R) plotted against
4. Applying the analytical result for product family development: redesign phase In the redesign phase, the results of the analysis of QFD and ISM illustrated in previous section are applied in product design and product family development. In this stage, four products were designed concurrently to satisfy requests of different markets. The design procedures are demonstrated in the following paragraphs. 4.1. Design for spatial variety The QFD approach listed in Table 1 indicates that Carafe (part no. 25) redesign is essential for addressing concerns regarding market variety. In Market 2, to maintain coffee temperature, the Carafe must use heat insulation material; moreover, the wall should be thickened, the shape of Carafe slenderized and the top narrowed to reduce heat loss. These changes of carafe specification also affect the design of other components. The incidence relation diagram of the Carafe is shown in Fig. 7 (extracted from Fig. 5). In this figure, the Carafe is the source component for the variant design (represented as a double-circle),
22
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 4 Incidence matrix after appropriate rearranging the order
Note: Grayed cells indicate the inter-chunk interactions.
exporting design constraints to the sink nodes. The design constraints are listed below: d(25, 21): The Heating Plate Module should fit the diameter of Carafe Base (fixed). d(25, 22): The Cup Bank should fit the diameter of Carafe body (changed). d(25, 24): The Carafe Handle should fit the arc and weight of Carafe body (changed). d(25, 26): The Carafe Cover should fit the diameter of Carafe rim, and the requested thermal condition (changed). d(25, 27): The Filter Module should fit the Carafe height (changed).
Fig. 7. Incidence diagram of Carafe (component 25).
The constraint of d(25, 21) is fixed (represented as dotted line in Fig. 7), and thus parts (20, 21) are
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
23
Under these constraints, the design of Filter Module (parts 27, 5, and 30) is changed from V- to Ushaped to fit the new Carafe design. Furthermore, constraint from the Filter Module is: d(5, 1): The Top Cover should fit the Basket Holder rim diameter. Since the specification of the Basket Holder rim is fixed, component 1 and 2 need not change their design. Consequently, Table 5 lists the variant component design driven by spatial market variety. 4.2. Design for temporal variety
Fig. 8. Constraint flow diagram of the Filter Module.
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. 7) owing to the variant carafe specification, resulting in the design of the Filter Module and Carafe Outfit Module having to be changed to match the altered conditions. In the Carafe Outfit Module, the components are redesigned to fit the new Carafe. However, the design change of the Filter Module must refer not only to the Carafe, but also to other components that provide constraints on the Filter Module, as shown in Fig. 8. Thus, in redesigning the Filter Module, the constraint from the Carafe becomes the source of variant design (represented as solid line in Fig. 8), while the others are fixed constraints (represented as dotted lines in Fig. 8) listed below: d(11, 27): The Housing should fit the Base. d(28, 30): The Filter Holder should fit the Filter Holder Packing Valve diameter. d(4, 5): The Top Cover Base should fit the Spout Seat diameter. d(3, 30): The Filter Holder should fit the Spout shape.
Table 2 indicates that the critical components for realizing temporal variety are the Housing (part 27), Top Cover (part 1), and Carafe (part 25). According to the hierarchical graph in Fig. 5, for these three components, the Carafe occupies the upper level in the interaction hierarchy. This arrangement means that the Carafe redesign should be addressed first, followed by that of the Housing and finally, the Cover. However, the incidence and costs involved in carafe redesign are quite high. The strategy of Company X thus is to ‘‘over-design’’ this component; that is, to improve the quality of the current specifications capable of handling future market requests. Therefore, the Carafe is upgraded for easy cleaning, pouring and dishwasher-safe, and both the ‘‘current’’ and ‘‘future’’ versions of the component are left unchanged. Therefore, according to the design priority, the product variety should focus on redesigning the Housing (part 27). To facilitate usability, the design team tends to substitute Swing-out Housing for the fixed housing. This change divides the component into two new parts; namely, the Swing-out Filter Housing and the Support. The Swing-out Filter Housing is further differentiated into either U- or V-shaped. The original design constraints of the Housing are laid on the two new parts, respectively (see Fig. 9). Thus, the shape of the Swing-out Housing thus must fit the Carafe; and the design of the Support must fit the Base. The variant design of the Housing directly influences the Top Cover (part 1); meanwhile, for convenient to use, the Top Cover is changed from a lift-up to a fixed design.
24
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 5 List of variant design driven by spatial variety Number
Component
Market 1
25 25* 22 22* 23 23* 24 24* 26 26* 27 27* 5 5* 30 30*
Glass carafe Thermal carafe Cup Bank of glass carafe Cup Bank of thermal carafe Handle Cover of glass carafe Handle Cover of thermal carafe Handle of glass carafe Handle of thermal carafe Cover of glass carafe Cover of thermal carafe V-shaped housing U-shaped housing V-shaped Cover Base U-shaped Cover Base V-shaped filter basket U-shaped filter basket
V
Market 2 V
V V V V V V V V V V V V V V
Note: *New design of the component.
Finally, Table 6 lists the variant design driven by temporal variety. 4.3. Developing a product family Four family products corresponding to different competitive conditions are derived based on the
variant designs developed above, the schematic diagram is illustrated in Fig. 10. Table 7 lists the components of the four products. Among these components, most of the variety occurs in chunks 1 and 4, while chunks 2 and 3 remain virtually unchanged, and thus are considered ‘‘platforms’’ of this product family. Moreover, the design team further suggested that components of the upper levels of chunk 3, including Water pipe Module, Heating Element, Base Module, and Heating Plate Module, should be standardized to reduce the redesign effort and production cost. A team of engineers and top managers of Company X estimated the sales volume, marketing, variable (raw material/production prices), and fixed (engineering/injection mold) costs for each product of using the product family designs, and compared these estimates
Table 6 List of variant design driven by temporal variety Number
Component
Current
27 27*1 27*2 1 1*
Housing Swing-out Filter Basket Support Lift-up Top Cover Fixed Top Cover
V
*
Fig. 9. Constraint flow diagram of the new design.
Future V V
V V *1
*2
Note: New design of the component. 27 and 27 are the two new designs of component 27.
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
25
Fig. 10. Schematic diagram of the final result.
Table 7 Components list of the product family Chunk
C1
Component
Number
Product 1
Product 2
Lift-up Top Cover Fixed Top Cover Top Cover set Spout Spout Seat Filter Holder Packing Valve V-shaped Cover Base U-shaped Cover Base V-shaped Filter Holder U-shaped Filter Holder V-shaped fixed housing U-shaped fixed housing V-shaped Swing-out Filter Housing U-shaped Swing-out Filter Housing
1 1* 2 3 4 28 5 5* 30 30* 27 27* 27*11 27*12
V
V
V V V V V
V V V V
Support Water Tank Cover
C2
Water Tank Silicone ring Water outlet pipe Pipe connection seat Packing Valve Heating Element Base Base Cover Switch Hot plate ring
C3
Hot plate Glass carafe Thermal carafe
Product 3
Product 4
V V V V V V
V V V V V
V V
V V
V
V
V V V V
*2
6
V
V
V V
V V
8 12 13 14 16 18 11 15 19 20
V V V V V V V V V V
V V V V V V V V V V
V V V V V V V V V V
V V V V V V V V V V
21 25 25*
V V
V
V V
V
27
V
V
26
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
Table 7 (Continued ) Chunk
Component
C4 *11
Note: 27
Number
Cup Bank of glass carafe Cup Bank of thermal carafe Handle Cover of glass carafe Handle Cover of thermal carafe Handle of glass carafe Handle of thermal carafe Cover of glass carafe
22 22* 23 23* 24 24* 26
Cover of thermal carafe
26*
*12
and 27
Product 1
Product 2
Product 3
V
Product 4
V V
V
V
V V
V
V
V V
V
V
V V
V
are new designs derived from component 27*1.
Table 8 Comparing costs and profit between independently developed and family product design % of current product
Sales volume Price VC FC MC Profit Total profit
No product family design
Product family design using this methodology
Product 1
Product 2
Product 3
Product 4
Product 1
Product 2
Product 3
Product 4
100 100 100 100 100 100
80 120 145 100 200 73
100 115 105 100 100 208
80 130 145 100 100 65
100 100 100 100 100 100
80 120 140 40 200 27
100 115 100 10 100 334
80 130 140 10 100 207
Current = 27
Future = 143
Current = 127
Future = 541
Note: VC, FC, and MC are the variable, fixed, and marketing costs, respectively.
to those for products designed independently. Table 8 lists the comparison. The profit is calculated using the following function: Pi ¼ Si ðPRi VCi Þ FCi MCi
only product launched. The result shows the potential savings and profit available using this methodology.
(1)
where Pi, Si, PRi, VCi, FCi, MCi are the profit, sales volume, price, variable cost, fixed cost, and marketing cost of product i, respectively. Table 8 illustrates that the primary cost difference between the two design strategies lies in the fixed cost. The product family designs significantly reduced the fixed cost for developing new products through sharing most components. The second row from the bottom shows that the profits associated with independently developing products 2 and 4 is minus 73% and 65% of current product, respectively. Therefore, the best decision seems to be not to develop any product in Market 2. However, product designs using this methodology generate a total profit 127% in current markets and 541% in future markets higher than if product 1 was the
5. Discussion Based on this investigation and application involving practical product design, for product family design through variety management, the hierarchical graph could support the following design strategies: (1) Design strategies for the dominant components: (a) The drivers of design variation of these components are customer requirements. Since the incidence and effort for the design variation are relative huge, this variation must be obvious to the customers, and customer requirements are likely to be considered as completely as possible. To achieve a stable Product Family Architecture,
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
‘‘over-design’’ of the components might be unavoidable in response to upgrade requests. (b) If the components are remained unchanged, they should be considered to be standardized or fixed specification, and become core platform of the product family. (2) The designs of the sink components are more likely to change to comply with both the altered design constraints and the requests of customer requirements. However, the cost and incidence of altering these components is relatively low. Furthermore, the redesign should incorporate the constraints provided by the source components. (3) Components that with their specification flow forms an interaction loop are likely to be further modularized or integrated.
6. Conclusions This study structurally illustrated the product family formation. The system employs two techniques: the ISM and QFD. Applications of these techniques to product architecture provide the hierarchical graph of interactions. Furthermore, the proposed methodology provides the following advantages for developing product family: (1) The methodology clarifies the specification flow between components rather than merely symmetric relationship ‘‘similarity’’ or ‘‘correlation’’. Thus, the necessary information is provided for determining not only clustering but also precedence among components. (2) The incidence matrix with the documented design constraints provides a computable way for design knowledge representation. (3) The hierarchical graphical diagram provides designers with a user-friendly display for clarifying the influence of each component variation. (4) The hierarchical structure along with the QFD analysis helps product family developers to identify whether the components should be standardized, altered, or modularized. (5) The integration of spatial and temporal market requirements is beneficial for establishing product family without redundant development costs and increasing product development lead time.
27
(6) The proposed approach can be integrated into a design information system (DIS) in the future for enhancing collaborative design and development.
References [1] B.J. Pine, Mass Customization: The New Frontier in Business Competition, Harvard Business School Press, Boston, 1993. [2] L. Cohen, Quality Function Deployment: How to Make QFD Work for You, Addison-Weesley, Reading, MA, 1995. [3] N.P. Suh, The Principles of Design, Oxford University Press, NY, 1990. [4] F.J. Erens, The synthesis of variety: developing product families, Dissertation Eindhoven University of Technology, 1996. [5] K. Fujita, K. Ishii, Task structuring toward computational approaches to product variety design, in: Proceedings of the ASME Design Engineering Technical Conferences, 1997, DETC97/DAC-3766. [6] M.V. Martin, K. Ishii, Design for variety: a methodology for understanding the costs of product proliferation, in: Proceedings of the ASME Design Engineering Technical Conference, 1996, 96-DETC/DTM-1610. [7] M.V. Martin, K. Ishii, Design for variety: development of complexity indices and design chart, in: Proceedings of the ASME Design Engineering Technical Conference, 1997, DETC97/DFM-4359. [8] M.V. Martin, K. Ishii, Design for variety: developing standardized and modularized product platform architectures, Research in Engineering Design 13 (2002) 213–235. [9] M.H. Meyer, J.M. Utterback, The product family and the dynamics of core capability, Sloan Management Review 34 (1993) 29–47. [10] M.S. Sawhney, Leveraged high-variety strategies: from portfolio thinking to platform thinking, Journal of the Academy of Marketing Science 26 (1998) 54–61. [11] K.T. Ulrich, S.D. Eppinger, Product Design and Development, second ed. McGraw-Hill, New York, 2000. [12] M.M. Tseng, J. Jiao, Design for mass customization by developing product family architecture, in: Proceedings of the ASME Design Engineering Technical Conference, 1998, DETC98/DTM-5717. [13] D. Robertson, K.T. Ulrich, Planning for product platforms, Sloan Management Review 39 (1998) 19–31. [14] K. Fujita, S. Akagi, T. Yoneda, M. Ishikawa, Simultaneous optimization of product family sharing system structure and configuration, in: Proceedings of the ASME Design Engineering Technical Conferences, 1998, DETC98/DFM-5722. [15] K. Fujita, H. Sakaguchi, S. Akagi, Product variety deployment and its optimization under modular architecture and module commonalization, in: Proceedings of the ASME Design Engineering Technical Conferences, 1999, DETC99/DFM-8923. [16] J.S. Yu, K.N. Otto, Product architecture definition based on customer demands, in: Proceedings of the ASME Int Design Engineering Technical Conferences, 1998, DETC98/DTM5679.
28
S.-W. Hsiao, E. Liu / Computers in Industry 56 (2005) 13–28
[17] T.W. Simpson, J.R.A. Maier, F. Mistree, A Product Platform Concept Exploration Method for Product Family Design, in: Proceedings of the ASME Design Engineering Technical Conference, 1999, DETC99/DTM-8761. [18] J.N. Warfield, Societal Systems: Planning, Policy and Complexity, Wiley, New York, 1976. [19] Y.U. Pimmler, S.D. Eppinger, Integration analysis of product decomposition, Proceedings of the ASME Design Theory and Methodology Conference 68 (1994) 343–351. [20] M.J.V. Wei, J.L. Greer, M.I. Campbell, R.B. Atone, K.L. Wood, Interfaces and product architecture, in: Proceedings of the ASME International Design Engineering Technical Conferences, 2001, DETC01/DTM-21689. [21] M.E. Sosa, S.D. Eppinger, C.M. Rowles, Designing modular and integrative systems, in: Proceedings of the ASME Design Engineering Technical Conference, 2000, DETC2000/DTM14571. [22] S.M. Salhieh, A.K. Kamrani, Macro level product development using design for modularity, Robotics and Computer Integrated Manufacturing 15 (1999) 319–329. [23] J. Hauser, D. Clausing, The house of quality, Harvard Business Review 66 (1988) 63–73. [24] J.N. Warfield, On arranging elements of a hierarchy in graphic form, IEEE-SMC 3 (1973) 121–132. [25] J.N. Warfield, A Science of Generic Design: Managing Complexity Through System Design, vol. 1, Intersystems, Salinas, CA, 1990. [26] J.N. Warfield, A Science of Generic Design: Managing Complexity Through System Design, vol. 2, Intersystems, Salinas, CA, 1990. [27] T. Kitamura, A computer-aided approach to the structural analysis and modification of a large circulatory system model, IEEE Transaction on Biomedical Engineering 46 (1999) 485– 493. [28] H.D. Sharma, A.D. Gupta, Sushil, The objectives of waste management in India: a futures inquiry, Technological Forecasting and Social Change 48 (1995) 285–309.
[29] F.S. Roberts, Discrete Mathematical Models, Prentice-Hall Englewood Cliffs, NJ, 1976. Shih Wen Hsiao is a professor in the Department of Industrial Design at the National Cheng Kung University, Tainan, Taiwan, Republic of China. He was also appointed to the Chairman of this Department in the period of 1998–2001. He received his PhD in mechanical engineering from National Cheng Kung University in 1990. Before joining the University in 1991, he spent about 13 years in the China Steel Corporation. His major research interests include fuzzy set theory, concurrent engineering, computer-aided design, neural network, genetic algorithm, gray theory, color planning for product design, heat transfer analysis, and application of reverse engineering in product design. He has published a lot of papers in the international journals in the above-mentioned areas. He was invited as a Keynote Speaker for the Conference on Grey System Theory and Its Applications, which was held in 2001. He got an Outstanding Research Award from Engineering College in 2002 and got a Distinguished Professor Award at National Cheng Kung University in 2003. Dr. Hsiao has been profiled in Who’s Who in the World, Who’s Who in Design, 500 Great Minds of the Early 21st Century, and 500 Distinguished Professors and Scholar of the BWW Society. Elim Liu is a PhD candidate in the Department of Industrial Design, National Cheng Kung University, Taiwan, Republic of China. She received her Bachelor degree and Master degree in Industrial Design from National Cheng Kung University in 1990 and 1993. Her research interests include product family development, and artificial intelligent based design inference system.