Proceedings of The 1996 ASME Design Engineering Technical Conferences and Computers in Engineering Conference August 18-22, 1996,

Irvine, California

96-DETC/DTM-1516
IMPLICATIONS OF MODULARITY ON PRODUCT DESIGN FOR THE LIFE CYCLE

Patrick J. Newcomb, Bert Bras, David W. Rosen Systems Realization Laboratory G.W. Woodruff School of Mechanical Engineering Georgia Institute of Technology Atlanta, Georgia 30332-0405 (404) 894-9668 Fax: (404) 894-9342 [email protected]

ABSTRACT Growing concern for the environment has spurred interest in environmentally conscious design and manufacturing. The concept of Design for the Life Cycle encompasses all aspects of a products life cycle from initial conceptual design, through normal product use, to the eventual disposal of the product. A products architecture, determined during the configuration design stage, plays a large role in determining the products life cycle characteristics. In this paper, modularity of product architectures with respect to life cycle concerns, not just product functionality and structure, is defined and applied in the analysis of architecture characteristics. A principal hypothesis underlying this work is that high degree of life cycle modularity can be beneficial across all viewpoints of interest because all interested people will view the product similarly and consistently. An architecture decomposition algorithm from the literature is adopted for partitioning architectures into modules from each life cycle viewpoint. Two measures of modularity are proposed: one that measures module correspondence between several viewpoints, and another that measures coupling between modules. The algorithm and measures are applied to the analysis and redesign of an automotive center console. Results of applying the algorithm and measures accurately reflected our intuitive understanding of the original center console design and predicted the results of our redesign. Furthermore, these measures incorporate only configuration

information of the product; hence, can be used before detailed design stages. 1 INTRODUCTION 1.1 Framework for Our Investigation Growing concern for the environment has spurred interest in environmentally conscious design and manufacturing (ECDM). The objective of ECDM is to minimize the negative environmental impacts of engineering systems. Although recycling has received a lot of recent attention in light of proposed European product take-back legislation, and while many think of recycling as the answer to our environmental problems, it is only a small part. A products whole life needs to be considered, from beginning to end. This cradle-to-grave design philosophy is generally termed Design for the Life Cycle (or Life-Cycle Design), and it encompasses all aspects of a products life cycle from initial conceptual design, through normal product use, to the eventual disposal of the product (Alting, 1993; EPA, 1993). Design for the Life Cycle encompasses so many things that it is necessary to narrow the field for discussion. This paper deals specifically with the aspects of configuration design concerned with Design for Assembly/Disassembly, Design for Service, and Design for the Post-Life, as well as design for the products intended functions. Our approach is to develop and analyze product architectures to satisfy requirements (as well as possible) from all of these life cycle viewpoints.

Copyright 1996 by ASME

Configuration design is concerned with determining what components are in a design and how they are arranged spatially and logically. Design of product architectures occurs during the configuration design stage, which is generally regarded as occurring after conceptual design but before parametric design (Dixon et al., 1988). Our first principal hypothesis is that for the majority of products, a products architecture plays the predominant role in determining its assembly, disassembly, recycling, service, and other post-life characteristics. Often, a products architecture is thought of in terms of its modules. For a car door, modules are its lock, window regulator, structural components, interior panel components, etc. A module is a physical or conceptual grouping of components. Modularity has been defined as the relationship between a products functional and physical structures such that (1) there is a one-to-one correspondence between the functional and physical structures, and (2) unintended interactions between modules are minimized (Ulrich & Tung, 1991). In this paper, the definition of modularity will be expanded to allow one-to-one correspondences between physical structures and structures of relevance to a life cycle viewpoint, similar to the Pahl and Beitz concept of production-oriented modules (Pahl & Beitz, 1988). Thus, modularity with respect to recycling can be defined, or with respect to servicing, etc. Such modularity will be termed life cycle modularity. Our second principal hypothesis is that high life cycle modularity can be beneficial across all viewpoints of interest. This implies that designers, manufacturers, recyclers, and maintenance personnel will tend to view a products structure in similar ways. For example, the door lock is a module from the designers viewpoint and would be a module from the recycling viewpoint if all of its components are made of compatible materials. Thus, a structural module of a product will correspond to a manufacturing module, a recycling module, and a service module. Such similarity in life cycle modules should be beneficial in that multiple views of the product are not necessary; the number of people involved in component and/or module design can be reduced; design groups can operate more independently; assembly, disassembly, and service costs can more easily be reduced. However, these principle hypotheses raise several questions. How can modules be defined in different viewpoints? How can modularity be measured across multiple viewpoints? This paper is concerned with the evaluation of design configurations with respect to function, service, and post-life issues. Our goal is to develop an evaluation tool that a designer can use in configuration design in order to

determine the degree to which a design simultaneously meets its function, service, and post-life goals in order to answer the questions raised above. Our approach towards this evaluation tool involves the use of product decomposition and module comparison to achieve product modularity. An existing decomposition algorithm is used. Measures of modularity are proposed to indicate the extent to which the product architecture achieves life cycle goals. A running example of an automotive center console is used to illustrate the application. The specific life cycle viewpoints to be investigated here include the following, with their defining characteristics: Material Recycling - material compatibility and separation effort; Service - servicing frequency: never, sometimes, frequent; Post-Life Intent - the intended destination for each component: incineration, material recycling, or component reuse. 1.2 Motivating Example: An Automotive Center Console An automotive center console often contains many features such as an armrest, a storage area, a cupholder, and vent controls (if back seats need air, as in a sedan) in order to perform many functions. Measured conventionally, a center console exhibits a fairly high modularity. Let us assume that a design team wants to improve an existing design of a center console (CC) with respect to life cycle issues including material recycling, component reuse, and service for a new vehicle release. Typically, a CC contains many different subassemblies, components and materials that complicate such a design task. The center console used as the example in this paper is from a 1993 Chrysler LHS and is shown in Figure 1. In Figure 1, modules which contain compatible materials (i.e., materials that can be recycled as a mixture and thus do not have to be separated before recycling) are highlighted. To be more specific, assume that the design team wants to reconfigure the CC by using modularity as a product architecture strategy. Essentially, the CC is a module by itself (of the car interior). However, there are many subassemblies and components contained within the console that can be intelligently reconfigured into modules by way of material choice, post-life intent, and spatial location. Not every subsystem or detail in the CC will be included in this design problem, but nineteen different console components will be used in the example. The goal of the design problem is to find a CC configuration with good post-life intent characteristics, good material recycling characteristics, and good service

characteristics, while maintaining the system functionality. By good post-life intent characteristics we mean that the modules and components designated for different post-life intents are easily separated from one another. For example, the components designated for reuse should be easily separable from the ones designated for material recycling. By good material recycling characteristics, we mean that the less separation into separate materials or compatible material groups, the better the design. The separation is dependent on two different aspects of the design: the materials of the components, and the actual physical connectivity between components since two components that are made of compatible materials need not be separated. Note that we assume a manual dismantling process. Our assumptions would differ for a mechanical separation process such as shredding (Coulter et al., 1996). Service considerations are similar in that separation ease is important, but only for those components that require servicing. If these components are grouped together, and access to them is easy, then the design should be considered a good one.

modularity. Using these modularity measures, it is possible to assess a designs architecture and to compare two or more designs with respect to life cycle issues. We assess the CC design in Section 4.0, then redesign it and compare these designs using our modularity measures. Advantages and limitations of our approach are also highlighted. 2.0 PRODUCT ARCHITECTURE AND MODULARITY Configuration design is the process of synthesizing product structures by determining what components and subassemblies are in the product and how they are connected and arranged spatially. Establishing the product structure involves the selection of modules and the design of module interfaces. Modularity is the concept of separating a system into independent parts or modules which can be treated as logical units. The way in which a product is divided into modules has a great affect on the way it is assembled, disassembled, serviced, and retired (among other things). The challenge, then, is to develop methods for configuration design that enable designers to meet requirements on modularity applied to all life cycle viewpoints of interest (assembly, service, recycling, reuse, etc.). Ulrich and Tung (1991) gave a summary of different types of modularity and their advantages and disadvantages. Benefits of modularity include simplified assembly and disassembly, improved serviceability, easier maintenance, and differential consumption. Potential costs of modularity include redundant physical architecture (due to decreased function sharing), excessive capability due to standardization (designing for the most rigorous application), and the potential for static product architectures and excessive product similarity. The authors also contend that modularity is a relative property that depends on two characteristics of a design: 1 Similarity between the physical and functional architecture of the design, 2 Minimization of incidental interactions between physical components. Accordingly, complete modularity is achieved when there is a one-to-one correspondence between the physical and functional architectures. This concept is used as the premise for modularity in this paper. The functional architecture of a product refers to the arrangement of a products overall function(s) into sub-functions. For example, the center console has many functions: hold cups, store items, and support arm (to name a few). The degree to which the physical architecture (the actual arrangement of components and modules) of a product resembles its functional structure contributes to modularity. In fact, in (Ulrich & Eppinger, 1995), the authors state that Perhaps the most important characteristic of a products architecture is its modularity. The modularity of

Mixed POM Vinyl Polystyrene

ABS

Steel Steel

Figure 1 Center Console (from a 1993 Chrysler LHS) So how does a designer go about finding a product configuration with good post-life intent characteristics, good material recycling characteristics, and good service characteristics, while maintaining the system functionality? How can the designer tell how good (or bad) the current design is with respect to theses issues? After a redesign is performed, how can the designer compare the two configurations to see which is better with respect to life cycle issues? These questions will be addressed throughout this paper. In the next section, we review different approaches to product modularity and configuration design. In Section 3.0, our product architecture decomposition algorithm is presented, along with our proposed measures of

a product influences its initial cost, ease of service (disassembly and reassembly), and effort required to retire the product. If a product must be disassembled to recycle its materials or reuse any of its components, the way in which these components are arranged into modules is of utmost importance. For example, car alternators are often recovered for remanufacturing after the car is retired. As a rule of thumb, the quicker and easier it is to remove the alternator from the car, the more economically feasible its remanufacture. If it were to take several hours to access and remove an alternator, few would pursue this task because it would not be worth the effort. Therefore, the fact that the alternator is a separate module that can be removed fairly easily (as opposed to being an integral part of the engine block) aids in both service and in product retirement. Life cycle design has received considerable attention in the literature. It is also promoted by the US Environmental Protection Agency (1993). We will only discuss selected works that highlight the role product architecture plays in life cycle issues. Ishii and his group (1995) successfully argued that product architecture strongly influences life cycle concerns. Their product structure representation scheme, their LINKER diagram, enables reasoning about product structures and drives several types of analyses, including serviceability. Their approach to clumping components into modules (Marks et al., 1993) helped motivate our study into a broader view of modularity. In this paper, their Semantic Links (attaches, connects, engages, supports, and covers) are used to model the connections between components. An additional viewpoint is that of the design of product families which Ishiis group is now investigating (Ishii et al., 1995). The German engineering standard VDI 2243 represents a good example of the state of the art in recycling and other life cycle concerns (VDI 1993). According to VDI 2243, the major problem area in the disassembly of all products lies in the separation of joints. This has provided the impetus for Beitz and coworkers to focus on knowledge-based systems to support fastener selection (Beitz 1993). For an industrial perspective, Xerox has been remanufacturing its copiers for several years. When copiers are returned from customers at the end of a lease, Xerox remanufactures them by reusing components and replacing worn components. They assemble remanufactured and new copiers on identical assembly lines with identical quality standards. Through the use of modular attachment methods and component standardization, Xerox has made their copies easier to disassemble, modify, and reassemble (Congress, 1992). With several life cycle viewpoints to be considered for a product with many components, determining the actual degree of correspondence between viewpoints is not easy to

do. Therefore, a systematic method is needed to handle this evaluation. Matrix Decomposition is used as our systematic method. Stewards Design Structure Matrix (DSM) (1981) model has been used to identify orderings of tasks and the organizationally difficult aspects of the design process (Smith et al., 1992). The DSM is a square matrix where the row elements and column elements are the design tasks to be performed. The matrix is binary; elements are filled based on whether or not one design task depends on information from another. The matrix is then partitioned into block-diagonal/lower triangular form with the blocks indicating coupled groups of tasks and the remaining entries below the diagonal indicating information transfer to later tasks. Rogers (1989) has developed software called DeMAID to aid in the organization of tasks. Eppinger has also used the DSM extensively in design organization, based upon product structure. In (Pimmler & Eppinger 1994), a method for the analysis of product design decompositions is detailed. The DSM partitioning method is used for two purposes in that paper: 1) organization of the development teams and 2) the definition of product architecture. This is accomplished by using components as the row and column elements (as opposed to design tasks in Stewards case). The method involves decomposition of the system into elements, the documentation of the interactions between the elements, and clustering (or integrating) the elements into architectural and team chunks. (Architectural chunks are equivalent to product modules.) The energy, material, information, and spatial (EMIS) interactions between elements are documented, and the DSM can be partitioned with respect to any of these viewpoints. In this paper, matrices that capture certain life cycle information are partitioned into clusters to indicate modular product architecture(s) from different life cycle viewpoints. Instead of using the EMIS interactions, this research uses the life cycle viewpoints stated at the end of Section 1.1. 3.0 PRODUCT ARCHITECTURE DECOMPOSITION AND MEASURES Corresponding to the two aspects of modularity presented in Section 2, two measures of modularity will be proposed in this section: a measure of how well modules from different viewpoints correspond, called the correspondence ratio (CR), and a measure of incidental interactions between modules, called cluster independence (CI). In order to measure the modularity of a product, it is first necessary to determine the products modules; to do so, we will describe the architecture decomposition algorithm used in our research. Some mathematical preliminaries will be covered first.

3.1 Set Notation Let X represent the set of components in a product or subassembly of interest. Components are grouped into modules because they share some property of interest that defines the module type. For example, material compatibility for recycling can define modules: all components that are compatible with one another are grouped into one module. A relation R matl can be defined that denotes material compatibility: x R matl y means that x and y are made of compatible materials (1) R matl is an equivalence relation that, when applied to a set of components, partitions those components into equivalence classes. The equivalence class under equivalence relation, R matl, of (component) c is denoted by [c]: [c] = {b | (b,c) in Rmatl } (2)

where b is any other element that is equvalent to c under relation R matl. In general, a module is defined by some equivalence relation applied to a set of components. For viewpoint i, the module that contains component x is denoted as V i(x) = [x] under the appropriate equivalence relation for that viewpoint. Equivalence relations for the other life cycle viewpoints are as follows: x R mrec y - x and y have the same material recycling characteristics; i.e., are of compatible materials and their separation difficulty is similar; x R pli y - x and y have the same post-life intent (shred/incinerate, recycle, reuse) x R serv y - x and y have the same service frequency (never, seldom, frequent) x R fun y - x and y implement the same function x R phy y - a designer perceives that structurally, x and y are part of the same structure. 3.2 Architecture Decomposition Product architecture decomposition into modules is performed by converting a matrix into block-diagonal or lower-triangular form. Other techniques exist for computing partitions of a set, but matrix decomposition has a beneficial visual interpretation. The algorithm developed by Kusiak and Chow (1987) was used since it is versatile enough to handle symmetric, asymmetric, and non-square matrices. This algorithm will be briefly explained using a material compatibility example. In the post life intent viewpoint, non-square matrices are used since each component is to be designated for one of four intents: recycling, reuse,

incineration, or landfilling. Therefore, the matrix would be 4 x m, where m is the number of components. In this work, a matrix containing material compatibility information is square, with the designs components as the row and column headings. Material compatibility measures are based on compatibility data and component volumes. The volumes are important because the compatibility of two materials sometimes varies based on which is the main ingredient and which is the additive. In this scheme, a 1 indicates good compatibility, a 2 indicates some compatibility, a 3 indicates limited compatibility, and a 4 indicates total incompatibility. The rule for decomposition of the material compatibility matrix is this: cluster the 1s (good compatibility measures) into blockdiagonal form. A more sophisticated rule that better reflects the 1-4 compatibility scale is under development, along with an improved decomposition algorithm. The algorithm consists of six main steps with iteration through all six. Figure 2 illustrates the process of decomposing the 19x19 matrix that describes material compatibility for the center console. Counter k is initialized to 1. Step 1. Select any row i of incidence matrix A (k) (A (k) denotes matrix A at iteration k) and draw a horizontal line through it. In Figure 2, a horizontal line is drawn through row 1. Step 2. For each entry 1 on the intersection with the horizontal line, draw a vertical line. In Figure 2, these vertical lines are drawn through columns A, B, and H. Step 3. For each entry 1 crossed by the vertical line, draw a horizontal line. In Figure 2, horizontal lines will be drawn through rows B and H.

Coverplate Armature Foam Pad Latch Top_hinge Bot_hinge Cup_fascia

A B C D E F G H

A 1 1 4 2 3 4 4 1

B 1 1 4 2 3 4 4 1

C 4 4 1 4 4 4 4 4

D 2 2 4 1 4 4 4 2

E 3 3 4 4 1 4 4 3

F 4 4 4 4 4 1 1 4

G 4 4 4 4 4 1 1 4

H 1 1 4 2 3 4 4 1

I 4 4 4 4 4 4 4 4

J 4 4 4 4 4 4 4 4

K 4 4 4 4 4 4 4 4

Figure 2 Fragment of matrix A(k) - Steps 1 and 2.

Step 4. Repeat Steps 2 and 3 until no crossed entries 1 remain. All double crossed entries 1 form a module. In this case, the module consists of components A, B, and H.

Step 5. Transform the incidence matrix A(k) into A(k+1) by removing the rows and columns corresponding to the horizontal and vertical lines drawn in steps 1 through 4. Step 6. If matrix A(k+1) = , stop; otherwise set k = k+1 and goto step 1. The resulting decomposition is shown in Figure 3. As can be seen, there are two main groups and eight individual components that are composed of unique materials.

when there is a one-to-one correspondence between the two architectures. The intersection of two modules does not tell us how well the modules correspond, however. Our measure of correspondence, the Correspondence Ratio (CR), is defined as:
CR = V i (x) V j (x) V i (x) V j (x)

(4)

A B HMO P Q S F G R N L K J I E DC Coverplate A 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 Armature EC fascia EC duct Left_base Bin Top_hinge Bot_hinge B 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 M1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 O 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 P 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 Cup fascia H 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4

where |X| indicates the cardinality (number of elements) of set X). If the correspondence between two modules is high, CR is close to 1, while if correspondence is low, CR is close to 0. The CR is a good measure on a module by module basis, but does not allow us to compare two designs very well since the different designs may have considerably different modules. A better evaluation of module correspondence for an entire product is the average CR for all modules in the product, called CRoverall:
CR overall =

Right_base Q 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 S 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 F 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4 G 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4

CR

(5)

#Modules

Fr_bracket R 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4 EC_louvers N 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 4 DC_power L 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 Ashtray

Bezel fascia

K 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 J 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 I 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 E 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 1 4 4 D 2 2 2 2 2 2 2 2 4 4 4 4 4 4 4 4 4 1 4 C 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1

Cupholder Latch Pad Foam

Figure 3 Decomposed material compatibility matrix.

3.3 Modularity Measures Since we are interested in the degree to which one viewpoints architecture corresponds to another viewpoints architecture, we compare the module sets from both architectures. To determine which components two modules have in common, the intersection of the modules can be computed: Vi(x) Vj(x) (3)

Similar to CR, CR overall = 0 indicates no correspondence between viewpoints. CR overall approaches unity when the individual module CRs approach unity. In order to address the second property of modularity, minimizing incidental interactions, a Cluster Independence (CI) measure has been developed. Minimizing incidental interactions is analogous to decreasing the amount of dependence between modules. For the purposes of design for the life cycle, the dependence between modules can be represented by the physical connections that exist between the modules. For example, for material recycling, connections within a module are not of greatest concern. Rather, connections between modules are important since they need to be broken in order to separate modules for recycling. Therefore, our second modularity metric depends upon the ratio of intra-module connections to the total number of connections in the product. Intra-module connections are those that are in blocks along the matrix diagonal of a decomposed matrix (e.g., Figure 3), referred to as on_block_diagonal_connections. CI is given in an informal manner: Cluster Independence
#on_block_diagonal_connections #total_connections

= (6)

where V i(x) denotes the module in viewpoint Vi that contains component x. Modularity is defined as the degree to which two architectures correspond. Complete modularity is reached

The metric is calculated using the on-block-diagonal measures because this yields the trend of zero being the worst and unity being the best for the metric. Theoretically, the maximum CI = 1.0 will occur only when all of the links are on the block-diagonal, meaning that no inter-module connections are in the product. That is, the product is a

disjoint set of modules, in which case it would probably be considered several products. The minimum CI is 0.0 and exists only in the case where each module consists of only one component. We propose an overall measure of modularity as the product of CRoverall and CI, which ranges between 0 and 1, inclusive (Eqn. 7). A weighted sum formulation could have been used, but we preferred the multiplicative form since its minimum value is 0. Modularity = (CRoverall)(CI) (7)

The material presented in this section will now be applied in the design of a center console. 4.0 CASE STUDY 4.1 Existing Design A Chrysler LHS center console has been chosen as an illustrative example due to its material mix and potential for improvement. The center consoles initial design was shown in Figure 1. In addition, a description of the design can be found in Table 1, along with some additional design requirements. The goal of the design problem is to find a CC configuration with good post-life intent (PLI) characteristics, good material recycling characteristics, and good service characteristics, while maintaining the system functionality. For the purposes of this example, we assumed certain post life and service intents.

The information shown in Table 1 can be used to determine the modules (sets of components that share a common characteristic) that exist from the different viewpoints. For instance, for material recycling, the ideal design would have all the components made of a single material since the product would not need disassembly and could just be reprocessed as is. As this is usually not possible, it is preferred that all components are made of materials compatible for recycling. Since this is not often realistic either, components with the same or compatible materials should be grouped together into modules. That is, compatible components should be connected to one another so that they form a structural module as well as a material compatibility module. The material compatibility matrix for the console was shown in Figures 2 and 3. Entries in the matrix were based on compatibility data and component volumes.

A B HMO P Q S F G R N L K J I E DC Coverplate A 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 Armature EC_fascia EC duct Left_base Bin B 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 M 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 O 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 P 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 S 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 F 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4 G 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4 Cup_fascia H 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4

Right_base Q 1 1 1 1 1 1 1 1 4 4 4 4 4 4 4 4 3 2 4 Top_hinge Bot_hinge

Table 1 Current Console Design.

Subassy Component coverplate armature foam pad latch top_hinge bot_hinge fascia cupholder fascia ashtray DC power fascia louvers ductwork left_base right_base front bracket bin Material ABS ABS PUR PVC POM steel steel ABS mix mix mix mix ABS PP ABS ABS ABS steel ABS Post Life Service Intent Intent recycle none reuse none reuse none incinerate none recycle frequent recycle sometime recycle sometime incinerate none incinerate frequent landfill none landfill none landfill sometime recycle none recycle frequent recycle none recycle none recycle none recycle none recycle none

Fr_bracket R 4 4 4 4 4 4 4 4 1 1 1 4 4 4 4 4 4 4 4 EC_louvers N 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 4 DC_power L 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 Ashtray

Bezel fascia

Armrest

K 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 J I 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4

Cupholder Latch Pad Foam

Cupholder

E 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 1 4 4 D 2 2 2 2 2 2 2 2 4 4 4 4 4 4 4 4 4 1 4 C 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1

Bezel

Endcap

Figure 4 Material Recycling (Compatibility and Separation) -- Original Design. As mentioned earlier, material recycling is also concerned with component disassembly in addition to material compatibility when one employs a manual

Base Bin

dismantling process. The less disassembly that has to be done to separate an assembly into material compatibility chunks, the better the design is for material recycling. Using the matrix format, this additional disassembly information can be shown along with the material compatibility. In Figure 4, the decomposed material compatibility matrix from the original design (see Figure 3) is augmented with physical connections between components shown by the shaded matrix elements. For each shaded element (i, j), the interpretation should be that the two components (i and j) are physically attached to one another. For example, in Figure 4, element (A, B) is shaded indicating that the Coverplate and Armature are physically attached to one another via some fastening method. Note that this matrix is also square and symmetric. This scheme is meant to show which components need to be disassembled from which others in order to separate the console into its material compatible modules. In a similar fashion, matrix decomposition can be used to obtain the life cycle viewpoints of post-life intent and service intent as well. Table 2 contains the module sets that result from the DSM decomposition for the three life cycle viewpoints on the original center console design. As can be seen in Table 2, the material recycling viewpoint has 12 different modules for 19 different components. This means that the console must be separated into 12 groups to perform material recycling on the whole console. A well known Design for Recycling rule of thumb is to reduce the number of different materials in a design in order to facilitate material recycling. However, what does it mean to reduce the number of material recycling clusters from 12 to 6, or from 12 to 1 for that matter? What is needed is a way of indicating how good or bad the separation effort is for a design. The assumption that improvement is made simply by reducing the number of modules neglects the actual physical connections between the components. Recall that (according to Ulrich and Tungs definition) one of the characteristics of modularity is the minimization of incidental interactions between components. This idea can be used to analyze the separation effort for a design. Just clustering a group of components together does not necessarily make the design better.

Table 2 Original Console Life Cycle Decomposition.

PLI with Connections cluster 1: cluster 1: armature bin coverplate fr_bracket cluster 2: rbase bin lbase rbase ec_ductwork lbase ec_louvers ec_ductwork ec_fascia ec_fascia bot_hinge cup_fascia top_hinge cluster 3: latch bot_hinge coverplate top_hinge cluster 2: cluster 4: fr_bracket foam cluster 5: armature ec_louvers cluster 3: pad cluster 6: dc_power cluster 4: cluster 7: ashtray cup_fascia cl. 8: bezel_fascia cupholder cluster 9: cupholder cluster 5: cluster 10: latch dc_power cluster 11: pad ashtray cluster 12: foam bezel_fascia #Modules = 12 #Modules = 5 Cluster Cluster Independence = Independence = 11/28 17/28 Material Recycling Service w/ Connections cluster 1: latch cl. 2: cupholder cl. 3: ec_louvers cluster 4: bot_hinge top_hinge cluster 5: dc_power cluster 6: pad foam armature coverplate cluster 7: bin fr_bracket rbase lbase ec_ductwork ec_fascia ashtray bezel_fascia cup_fascia #Modules = 7 Cluster Independence = 20/28

The actual arrangement and physical connectivity between components can be considered by including the Cluster Independence measure. CI for material recycling was determined by first counting the number of unique (symmetric matrix) shaded squares within the blocks along the diagonal in Figure 4, while ignoring diagonal entries (e.g., A-A). In this case, there are 11 shaded entries representing 11 intra-module connections. The total number of shaded entries is 28, again ignoring diagonal and redundant entries. Hence, the CI measure is 11/28. As a general rule, CI penalizes architectures with many small modules, even though, at first glance, such architectures might be considered highly modular. Hence, the CI value for material recycling is smaller than for post life intent and servicing viewpoints. Table 3 shows some of the modules that result from comparing viewpoints. For example, in the first column, the Material Recycling and PLI Separation viewpoints are compared, yielding 14 different modules. The modules are listed with each modules Correspondence Ratio (CR) following a list of the modules components. For example, the first module contains only the component Coverplate and has a CR = 0.083. This CR was calculated by applying Eqn.

4 to Material Recycling Cluster 1 and PLI Cluster 1 from Table 2. In this case, V i (x) V j (x) = {coverplate} which contains just one component, while V i (x) V j (x) contains 12 components, yielding a CR of 1:12 = 0.083. As can be seen in Table 3, the original console has relatively low Correspondence Ratios, except for the Material Recycling with Service case (CRoverall = 0.583) which accounts for the many one-component modules it contains. The on_block_diagonal_ connections are pretty low for them all however, indicating a higher separation effort. Table 3 Original Console Life Cycle Viewpoint Correspondence
Material Recycling, PLI Separation coverplate CR = 1:12 = 0.083 armature CR = 1:3 = 0.333 bin rbase lbase ec_ductwork ec_fascia CR = 5:12 = 0.417 bot_hinge top_hinge CR = 2:11 = 0.182 ... #Modules = 14 SUM_CR = 4.431 CR_overall = 0.316 Number of on_block_diag connections = 7 Material Recycling, Service Separation coverplate armature CR = 2:4 = 0.500 bin rbase lbase ec_ductwork ec_fascia cup_fascia CR = 6:9 = 0.667 bot_hinge top_hinge CR = 2:2 = 1.000 ... #Modules = 12 SUM_CR = 7.000 CR_overall = 0.583 Number on_block_diag connections = 11 All Viewpoints coverplate CR = 1:14 = 0.071 armature CR = 1:4 = 0.250 bin rbase lbase ec_ductwork ec_fascia CR = 5:14 = 0.357 bot_hinge top_hinge CR = 2:11 = 0.182 ... #Modules = 14 SUM_CR = 2.747 CR_overall = 0.196 Number of on_block_diag connections = 7

product of the CRoverall and CI measures for each viewpoint comparison of interest. Table 4 summarizes the results of the modularity evaluation for the original center console design, with Modularity ranging from 0.051 (for the intersection of the Service, PLI, and Material Recycling viewpoints) to 0.238 (for the intersection of the Material Recycling and Service viewpoints). It makes sense that the comparison of all three viewpoints would yield the lowest Modularity measure because less correspondence between viewpoints is likely as more viewpoints are added. Since it is known that unity is the theoretical best for the modularity metric, does it follow that the results obtained here indicate poor modularity, or is this as good as modularity gets? This can be investigated by performing an intelligent redesign on the center console that is intended to raise its modularity. The redesigned CC can then be re-evaluated and compared to the original to see if the measures allow a designer to predict the results of design changes. 4.2 Redesign of the Center Console What are some changes that would intuitively improve the CCs modularity with respect to all life cycle viewpoints? Reducing the number of materials, intelligently clustering compatible materials, and rethinking the post-life intent and service needs could all be beneficial to the modularity assessment and will be pursued here. For example, the complicated retractable cupholder design (which is made up of a mixture of plastic and metal components) could be changed to a single piece of molded ABS plastic that simply attaches to the same spots. Notice that in addition to improving the material recyclability of this component, this sort of configuration change affects the expected serviceability. Whereas the complicated original design is expected to have a frequent need for service due to the many moving parts, the new design should require no service. Changing the Post Life Intent of the cupholder assembly from incineration to recycling also helps this component fit into a module with other components slated for recycling. Other simple changes, such as making the Latch and Front_bracket out of ABS instead of POM and steel, respectively, making both the Foam and Pad out of polyurethane foam (PUR), and making the Bezel entirely out of ABS instead of a mix of materials were also made. Changing the post-life intent for the Pad from incineration to reuse provides a couple of advantages in this modularity evaluation: 1) it rounds out the elimination of the incineration category, helping to tighten the correspondence between viewpoints, and 2) it further groups the Pad with the Foam component, as they have been configured with the same material as well. A summary of the redesigned console is given in Table 5.

Table 4 Summary of Original Console Design

# Modules Matl Recycle Post Life Intent Service w/ connections Mat_rec, PLI PLI, Service Mat_recycling Service All Viewpoints 12 5 7 14 11 12 14 CR N/A N/A N/A 0.316 0.250 0.583 0.196 CI 11/28 17/28 20/28 7/28 11/28 11/28 7/28 Modularity N/A N/A N/A .082 .102 .238 .051

The final measure used for the modularity evaluation is the Modularity metric which, as noted in Section 3, is the

Now that the CC has been redesigned, it can be compared to the original design to determine how well the measures reflect the changes. Table 6 shows the new console configuration with only 5 material recycling clusters, as opposed to 12 in the original. Also, the number of service clusters reduced to 7 from 5, while the Post-life Intent clusters were shuffled around, but remained at a total of 4. We indicated that it was important to reduce the effort that goes into separation activities, whether for recycling, other product retirement, or service. We have reduced the separation effort for the material recycling and PLI viewpoints, whereas the Service stayed the same as indicated by the CI measures. For space considerations, a table of viewpoint correspondences is omitted. However, in the comparison given in Table 7, results of viewpoint comparisons are given. It is evident that the number of modules decreased significantly with the redesign. Although this is not necessarily always a positive trend, note that, in general, a reduction in the number of modules yields a reduction in the separation effort required. The redesign illustrates this point: it has fewer modules and better CI values. While the smaller number of modules shows a trend towards better individual Life Cycle viewpoints, the CI values show a better modular design for the combined Life Cycle viewpoints, as is also indicated by the increase in the CRs. Sometimes the higher CI measures actually follow the high CRs because the modules become so large that they envelop the previously off-block-diagonal connections. Table 5 Redesigned Center Console.
Subassy Comp. Material Post Life Intent recycle recycle reuse reuse recycle recycle recycle recycle recycle recycle landfill recycle recycle recycle recycle recycle recycle recycle Service Intent

Table 6 Redesigned Console Life Cycle Decomposition.

Service w/ Connections cluster 1: cluster 1: latch latch armature cluster 2: coverplate ec_louvers cluster 2: cluster 3: bin bot_hinge fr_bracket top_hinge rbase cluster 4: lbase dc_power ec_ductwork cluster 5: ec_louvers pad ec_fascia foam ashtray armature bezel_fascia coverplate cupholder cluster 6: cluster 3: bin dc_power fr_bracket cluster 4: rbase bot_hinge lbase top_hinge ec_ductwork cluster 5: ec_fascia pad ashtray foam bezel_fascia cupholder #Modules = 5 #Modules = 3 #Modules = 6 Cluster Independence Cluster Independence Cluster Independence = 20/26 = 23/26 = 20/26 Material Recycling PLI with Connections cluster 1: bin fr_bracket rbase lbase ec_ductwork ec_louvers ec_fascia ashtray bezel_fascia cupholder bot_hinge top_hinge latch armature coverplate cluster 2: pad foam cluster 3: dc_power

Table 7 Comparison of Console Designs.

none none none none frequent sometimes sometimes none none none sometimes none frequent none none none none none # Modules Material Recycling Post-Life Intent Service w/ conn. Matl_rec, PLI PLI, Service Mat_rec, Service All VPs Original 12 5 7 14 11 12 14 Redesign 5 3 6 5 7 7 7 CRoverall Original dna dna dna .316 .250 .583 .196 Redesign dna dna dna .600 .355 .605 .355

Armrest

Cupholder Bezel

Endcap

Base Bin

coverplate armature foam pad latch top_hinge bot_hinge cupholder bezel_fascia ashtray DC power fascia louvers ductwork left_base right_base front bracket bin

ABS ABS PUR PUR ABS steel steel ABS ABS PC mix ABS ABS ABS ABS ABS ABS ABS

10

Material Recycling Post-Life Intent Service w/ conn. Matl_rec, PLI PLI, Service Mat_rec, Service All VPs

Cluster Independence Original Redesign 11/28 20/26 17/28 20/28 7/28 11/28 11/28 7/28 23/26 20/26 20/26 18/26 18/26 18/26

Modularity Original dna dna dna .082 .102 .238 .051 Redesign dna dna dna .444 .246 .419 .246

The real key here is the Modularity measure that depends upon both the correspondence and independence aspects of modularity. As can be seen in Table 7, overall Modularity increased nearly 5 times for all viewpoints, from 0.051 to 0.246, as indicated in the last row. All other Modularity measures increased significantly. Increases in Cluster Independence drove much of the Modularity increase. In particular, notice the improvement in Material Recycling/Post-Life Intent CI from 7/28 to 20/26. This indicates that connections between components of compatible materials increased significantly, contributing to the formation of common modules across multiple viewpoints. These measures indicate that there was room for improvement in the individual life cycle viewpoints and in the combination of the viewpoints. More importantly, the positive effects of some simple configuration changes were predictable before the redesign was evaluated and verified by its evaluation. While the Modularity measures for the redesign are still not close to unity, they are significantly better than the original console designs measures. This experiment helps to provide support for the usefulness and correctness of the metrics. 5.0 CONCLUSIONS The application of product modularity to design for the life cycle practices has been explored in this paper. Specifically, modularity with respect to life cycle viewpoints, not just product functionality and structure, was defined and applied in the analysis of product architecture characteristics. An algorithm to partition an architecture into modules for a given viewpoint was presented. Two measures of modularity were proposed to measure module correspondence between several viewpoints, and to measure coupling between modules. An automotive center console was analyzed and redesigned based upon the results of applying the measures.

Results of this study are summarized as follows: the concept of life cycle modularity leads to a uniform method of reasoning with the implications of life cycle requirements on a products architecture; the proposed measures accurately reflected our intuitive understanding of the original center console design and predicted the results of our redesign; these measures incorporate only configuration information of the product; hence, can be used before parametric and detailed design stages; the decomposition algorithm works effectively for asymmetric and non-square matrices. The results of this preliminary study are promising. A more complicated example of an automotive interior is currently underway to better exercise the decomposition algorithm and modularity measures. Other current work includes developing modularity measures for later design stages that consider more detailed information, deriving specific redesign suggestions from the modularity measure results, and integrating the algorithm and measures to our CAD system for architecture analyses directly from assembly models of products. Further collaboration with industrial sponsors will better ground the concepts of life cycle modularity in design practice. ACKNOWLEDGMENTS Financial support from NSF, Grant No. DMI-9414715, is gratefully acknowledged. We also acknowledge the Chrysler Corporation for their donations of automotive components. REFERENCES Alting, L., 1993, Life-Cycle Design of Products: A New Opportunity for Manufacturing Enterprises in Concurrent Engineering - Automation, Tools and Techniques, A. Kusiak Ed., John Wiley and Sons, New York, pp. 1-17. Beitz, W, 1993, Designing for Ease of Recycling - General Approach and Industrial Applications, Proceedings 9th International Conference on Engineering Design, Zurich, Aug. 17-9, pp. 731-8. Congress, 1992, Green Products by Design: Choices for a Cleaner Environment, OTA-E-541, Office of Technology Assessment, Washington, DC. Coulter, S L, Bras, B A, Winslow, G and Yester, S, 1996, "Designing for Material Separation: Lessons from the Automotive Recycling, ASME Design for Manufacturing Symposium, ASME Design Technical Conferences, Irvine, California, August 22-24. Dixon, J R, Duffey, M R, Irani, R, Meunier, K, and Orelup, M, 1988, A Proposed Taxonomy of Mechanical Design

11

Problems, ASME Computers in Engineering Conference, pp. 41-46. EPA, U.S., 1993, Life-Cycle Design Guidance Manual, EPA/600/R-92/226, US Environmental Protection Agency, Office of Research and Development, Washington DC. Ishii, K, 1995, Life-Cycle Engineering Design, ASME J. of Mechanical Design, Vol. 117(B), pp. 42-7. Ishii, K, Juengel, C, and Eubanks, C F, 1995, Design for Product Variety: Key to Product Line Structuring, Proceedings ASME Design Theory and Methodology Conference, DE-Vol. 83, Boston, pp. 499-506. Kusiak A and Chow, W S, 1987, Efficient Solving of the Group Technology Problem, J. of Manufacturing Systems, 6(2), pp. 117-124. Marks, M D, Eubanks, C F and Ishii, K, 1993, Life-Cycle Clumping of Product Designs for Ownership and Retirement, Proceedings ASME Design Theory and Methodology Conference, T.K. Hight and L.A. Stauffer Ed., Albuquerque, New Mexico, pp. 83-90. Pahl, G. and Beitz, W., 1986, Engineering Design. SpringerVerlag, London/Berlin. Pimmler, T U and Eppinger, S D, 1994, Integration Analysis of Product Decompositions, Proceedings ASME Design Theory and Methodology Conference, DE-Vol. 68, pp. 343-351. Rogers, J.L., 1989, A Knowledge-Based Tool for Multilevel Decomposition of a Complex Design Problem, NASA Technical Paper 2903. Rosen, D W, Bras, B A, Hassenzahl, S L, Newcomb, P J, Yu, T, 1994, Computer-Aided Configuration Design for the Life Cycle. Accepted in: Journal of Intelligent Manufacturing. Smith, R P, Eppinger, S D, and Gopal, A, 1992, Testing an Engineering Design Iteration Model in an Experimental Setting, Proceedings ASME Design Theory and Methodology Conference, DE-Vol. 42, pp. 267-276. Steward, D V, 1981, Systems Analysis and Management: Structure, Strategy and Design, Petrocelli Books, New York. Ulrich, K T and Tung, K, 1991, Fundamentals of Product Modularity, Proceedings ASME Winter Annual Meeting Conference, DE Vol. 39, Atlanta, pp. 73-80. Ulrich, K T and Eppinger, S D, 1995, Product Design and Development, McGraw-Hill, New York. VDI, 1993, Konstruieren Recyclinggerechter Technischer Produkte (Designing Technical Products for Ease of

Recycling), VDI 2243, VDI-Gesellschaft Entwicklung Konstruktion Vertrieb, Germany.

12