Non-Functional Influences on Product Architecture by Jeffrey B. Dahmus B.S. Mechanical Engineering Stanford University, 1997 Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the Requirements for the Degree of Master of Science at the Massachusetts Institute of Technology June 2001 BARKER C 2001 Massachusetts Institute of Technology All rights reserved All ight resrvedMASSA CHUSETTS INSTITUTE OF TECHNOLOGY JUL 16 2001 LIBRARIES ........................ Department of Mechanical Engineering Signature of A uthor.............. May 11, 2001 .............................................. C ertified by ............................. Kevin N. Otto Associate Professor of Mechanical Engineering Thesis Supervisor A ccepted by ............................................................ ........................................................................... Ain A. Sonin Professor of Mechanical Engineering Chairman, Department Committee on Graduate Studies 2 Non-Functional Influences on Product Architecture by Jeffrey B. Dahmus Submitted to the Department of Mechanical Engineering on May 11, 2001 in Partial Fulfillment of the Requirements for the Degree of Master of Science ABSTRACT Determining product architecture is a key step in any product development activity. By selecting an effective product architecture for a product, companies can realize large savings, both during development and over the life of the product. This thesis explores how non-functional issues such as serviceability, technology change rate, and supply chain can and should be considered when determining product architecture. Each of these influences is carefully examined, and detailed methodologies are developed for incorporating such concerns into product architecture decisions. These methodologies involve first determining functionally based modules through the application of product modularization heuristics. These modules can then be evaluated from the perspective of a given influence by using methods and equations presented in this thesis. Application of such methods helps to determine if the suggested modules are effective from the standpoint of a particular influence. The methodologies developed in this thesis allow a product development team to architect a product while taking into account serviceability, technology change rate, and supply chain influences. By taking such concerns into consideration, product architectures can be developed that are effective for the product and company. Thesis Supervisor: Kevin N. Otto Title: Associate Professor of Mechanical Engineering 3 4 Acknowledgements I would like to thank my advisor, Kevin Otto, for his support and guidance over the past two years. His insights, advice, optimism, and humor have helped to make my experience a productive and enjoyable one. I am very grateful that I had the opportunity to work with him. I would also like to thank Michael Furst, Stephen Hoover, Abu Islam, and Gregory Kott of Xerox for providing numerous insights into actual product development processes. Their advice, comments, and feedback helped to bridge the gap between research and practice. I would also like to thank the MIT Center for Innovation in Product Development and the Office of Naval Research for their sponsorship of my research. Lastly, I would like to thank my friends and family for their continued encouragement and support. 5 6 Table of Contents S Introduction.......................................................................................................................................... The Importance of ProductArchitecture....................................................................................... 1.1 1.2 DeterminingProductArchitecture ................................................................................................ 1.3 Scope of this Thesis ......................................................................................................................... 1.4 Organizationof this Thesis .............................................................................................................. 2 Product Architecture......................................................................................................................... 2.1 Terminology..................................................................................................................................... 2.2 ManagingProductArchitectures .................................................................................................. 2.3 3 4 Determining ProductArchitectures.............................................................................................. 9 11 13 15 15 17 17 18 19 Functional M odeling.......................................................................................................................... 21 3. 1 Background..................................................................................................................................... 22 3.2 Function Structures......................................................................................................................... 22 3.3 Modularity Heuristics...................................................................................................................... 26 Serviceability ...................................................................................................................................... 29 Related Work ................................................................................................................................... 29 4.2 Approach ......................................................................................................................................... 31 4.1 Modeling Serviceability................................................................................................................... 33 4.3.1 Basic Service Cost Equations .............................................................................................. 33 4.3.2 Basic Reliability Concepts................................................................................................... 34 4.3.3 Random Failure M odel........................................................................................................... 37 4.3.4 Periodic M aintenance M odel.............................................................................................. 42 4.4 Example: Xerox Document System .............................................................................................. 47 4.4.1 Charging M odule .................................................................................................................... 49 4.4.2 Cleaning M odule..................................................................................................................... 54 4.3 4.5 Discussion ....................................................................................................................................... 7 56 5 Technology Change Rate ................................................................................................................... 59 5.1 Related Work ................................................................................................................................... 60 5.2 Approach ......................................................................................................................................... 62 5.3 Technology Change Rate Values ..................................................................................................... 63 5.3.1 When W ill the Technology Change? ...................................................................................... 63 5.3.2 Secondary Questions ............................................................................................................... 65 5.3.3 Using Technology Change Rate Values ................................................................................. 70 5.4 Example: Black and Decker Screwdriver........................................................................................ 71 5.4.1 5.5 Example: Xerox Document System .................................................................................................. 76 5.5.1 5.6 6 8 Charging M odule .................................................................................................................... 77 Discussion ....................................................................................................................................... 81 Supply Chain ...................................................................................................................................... 83 61 Related Work ................................................................................................................................... 83 62 Approach......................................................................................................................................... 85 63 EvaluatingSuppliers ....................................................................................................................... 86 64 Example: Black and Decker Screwdriver........................................................................................ 89 6.4.1 Torque M odule ....................................................................................................................... 91 6.4.2 Power M odule ......................................................................................................................... 93 65 7 Power M odule ......................................................................................................................... 72 Discussion ....................................................................................................................................... 95 Conclusions ......................................................................................................................................... 97 7.1 Optimization.................................................................................................................................... 97 7.2 Future Work................................................................................................................................... 105 Notes .................................................................................................................................................. 107 I Introduction The success of a company is rarely defined by the success of a single product. Instead, long-term company success and market growth often depend on providing the customer with a broad and continuous stream of value-rich products. One does not need to look far to find examples. Intel, in the 20 years since its first microprocessor was launched, has released over a dozen more microprocessor generations.' With each passing product generation, Intel's position as one of the most profitable and influential players in the computer industry has been further strengthened. Looking to the future, more generations of microprocessors are certainly in the works. In America, Honda has developed a solid reputation for producing dependable, high-quality automobiles. Customers interested in the Honda Accord sedan, one of the mainstays of the Honda line, can choose from five different models, ranging from the basic DX model to the fully equipped EX V6.2 This wide range of offerings helps the Accord reach numerous different market segments. Sony, often regarded as one of the premier developers of electronic goods, produces a broad and almost-constantly evolving line of products. Its very successful Walkman line has seen literally hundreds, perhaps thousands, of different iterations since its initial introduction in Japan in 1979.3 Through this method of product development, Sony has dominated the personal portable stereo market, with products designed for virtually every market niche. Intel's change-intensive approach to new product development confers considerable advantages. By rapidly introducing new technologies, Intel can be more responsive to customer needs, providing the products and performance that customers demand when they demand it.4 Firms like Intel, which compete in fast clockspeed industries, must continually push the technology envelope, keeping products innovative and new.' Intel releases a steady stream of new products, each featuring the newest technological improvements, from faster clock speeds to enhanced graphics capabilities. Because of this approach, product developers at Intel do not need to delay products until every last new feature is fully developed. Instead, new features that are not incorporated into the latest product, will simply be included in the next iteration. In a product development environment such as this, the main concern is simply that the new model be better than the previous one. 6 With incremental improvements, companies can also limit their risk, as new products often do not represent major expenditures of company time and resources. While 9 there are certainly many reasons for Intel's success, part of it can be attributed to its steady introduction of new products. Honda's variety-intensive approach also provides the company with a competitive advantage. With a range of product offerings, Honda can better tailor its Accord models to the needs of different market segments. In doing so, each product can address a specific market niche, thereby allowing Honda to gain market share and increase profitability. Satisfying the diverse needs of a large customer pool, while difficult, is often necessary to provide sufficient value to customers. Honda's method of offering variety can be directly contrasted to Henry Ford's approach with the Model T. The Model T offered the customer no variety, as evidenced by Henry Ford's famous comment, "You can have any color car you want as long as it's black." 7 Although many factors contributed to the eventual decline in Model T sales, one important factor involved the shortcomings of the single model philosophy. In 1925, with General Motors now offering "a car for every purpose and every purse," Ford's one product approach was no longer effective in the marketplace. 8 Automobile consumers now wanted a wider range of product offerings and more frequent model changes. Some firms, such as Sony, actively practice both change-intensive and variety-intensive product development. In what is sometimes called "hyper-variety," Sony produces hundreds of variants of each product over a short period of time. 9 For example, with the popular Sony Walkman, Sony introduced over 250 different models to the US market during the 1980's.O With such prolific new product development, Sony was able to produce a Walkman model for almost every imaginable market niche. With so many different product offerings, Sony's approach to product development could almost be thought of as mass customization, not mass production." Not surprisingly, Sony dominated the personal portable stereo market in the US, capturing 45-50% of market revenue in 1990.12 With this broad and rapid product development process, Sony has developed a reputation as one of the most consistent and impressive developers of consumer and industrial electronics.13 Intel, Honda, and Sony are good examples in today's marketplace of companies striving to decrease time to market and increase product line breadth. By decreasing time to market, companies are able to produce products quickly, and are thus better able to actively respond to shifting market needs. By offering a wide variety of products, companies can better satisfy the diverse needs of the market. By doing both, companies can often dominate their industries. 10 1.1 The Importance of ProductArchitecture Given these distinct advantages to maintaining broad and constantly updated product lines, companies have looked for ways to offer such variety at reduced costs to the company. One such method that has proven successful involves the use of carefully determined product architectures. Product architecture, as defined by Ulrich, is the scheme by which the function of a product is allocated to physical components. 4 Choosing an appropriate product architecture is a key step in any product development process as it can have an enormous impact on the success of a product. In turn, companies also have a great deal of latitude in selecting product architectures and must weigh many often-competing objectives when attempting to select an appropriate architecture. How important is the selection of an appropriate product architecture? While some industries have only recently begun to realize the value of selecting a suitable product architecture, other industries, notably the automobile industry, have used this idea for quite some time. It is common to hear about automobile companies basing new cars on existing platforms, thereby allowing companies to release cars faster and at reduced costs. The term product platform, as defined by Robertson and Ulrich, is the collection of assets, ranging from actual components to knowledge, which are shared by a set of products.' 5 In the case of the auto industry, what constitutes a platform differs greatly between different carmakers. Among automobile companies, Volkswagen (VW) is generally considered to be the leader in product platforming. In 1999, its A4 platform was used to produce over 1.9 million cars, once again making it the world's best-selling platform.' 6 This confusingly named platform provides the basis for seven distinct car lines: the VW Golf, VW Beetle, VW Bora, Audi A3, Audi TT, Skoda Oktavia, and Seat Toledo." In the case of VW, the platform includes front axles, rear axles, wheels, steering systems, front ends, rear ends, exhaust systems, seat frames, wiring, brake systems, fuel tanks, and more.'1 VW claims that approximately 65 percent of the components in the VW Golf are shared by the rest of the cars built on the same platform.'" While this represents a great deal of sharing, VW is striving to reach 70 percent commonality between this family of cars. 0 11 Figure 1.1: Fleet of Volkswagen cars produced from the A4 platform. On the left is the VW Golf. On the right (from top to bottom) are the VW Beetle, Audi TT, VW Bora, Audi A3, and Seat Toledo. In adopting this strategy of product platforming, VW cites numerous benefits. Perhaps the most obvious is the sizeable reduction in cost, both in development and in production. In 1999, VW 22 Through claimed to save $1.7 billion annually through its product platforming strategy. platforming, VW was also able to reduce new model development time, lower component prices, and reduce component counts within the group. Improved product quality was also realized by VW, as technologies already tested in one car model could be successfully introduced into other models built on the same platform. Perhaps the greatest risk in using a platform-based product strategy is that all products may end up being too similar. VW must make sure that the owner of an Audi TT feels as if he is driving a sporty coupe, not a small family sedan such as the VW Bora. VW prevents this partially through styling cues, which are clearly quite different between the various automobiles. VW also differentiates the cars by changing the height of the driving position and by modifying the suspension characteristics. By simply altering these few properties, VW claims that the cars can be sufficiently differentiated in the eyes of the customer. Other carmakers also actively pursue the idea of using product platforms to reduce costs. In February of 2001, Daimler-Chrysler announced disappointing results for the fourth quarter of 2000. The world's fifth-largest automaker reported loses of $269 million for the October through December period of 2000. This came in stark contrast to a profit of $1.5 billion for the same period of 1999.25 In starting to address these disappointing financial numbers, 12 DaimlerChrysler CEO Juergen Schrempp and Chrysler CEO Dieter Zetsche introduced a plan for recovery. The plan featured four main components: eliminating 26,000 jobs over the next two years, demanding immediate price cuts from suppliers, pushing ahead with the introduction of new Chrysler vehicles, and sharing more of the underlying auto components between Mercedes, Chrysler, and Mitsubishi.2 6 The mere fact that DaimlerChrysler would include the idea of product platforms as a main point in their recovery, demonstrates the cost-cutting value of such techniques. In their plan for additional component sharing among their automobiles, DaimlerChrysler listed numerous areas for greater sharing, including electronic controls, monitoring systems, seat frames, rear axles, transmissions, and steering systems. It also suggested sharing engines, placing some Mercedes diesel engines into Chrysler cars such as the PT Cruiser and Jeeps, although only in those versions sold in the European market.2 8 As far as automobile platforms, DaimlerChrysler suggested much greater platform sharing between Chrysler and Mitsubishi, reducing the total number of platforms used by DaimlerChrysler from the current 29, down to 12 to 16.29 Currently, Chrysler already uses a Mitsubishi platform as the basis for its Chrysler Sebring and Dodge Stratus sedans. While much greater sharing is scheduled at DaimlerChrysler, company insiders are still hesitant to leverage platforms too much, for fear of blurring the distinction between Mercedes and Chrysler automobiles. DaimlerChysler executives have religiously defended the Mercedes brand, trying to maintain its lofted brand image as a fine luxury car. In fact, Mercedes executives have repeatedly commented that protecting the Mercedes brand is the utmost priority for the company. 31 Mercedes CEO Juergen Hubbert made it clear that there would be no sharing of components beyond those that are largely invisible to the typical car buyer. As for using Mercedes platforms in Chrysler's American automobiles, such an idea was dismissed by Hubbert as not being a topic of discussion. 3 1.2 Determining ProductArchitecture Both the Volkswagen and DaimlerChrysler examples show some of the opportunities and difficulties associated with product platforms. While enormous cost savings can be realized, the risk that products will no longer be differentiated in the eyes of the customer is very real. While in theory, product platforms provide a straightforward way for companies to reduce product 13 development time and cost, in practice, making actual product platform decisions is quite difficult. One difficulty lies in the fact that determining product architecture is not a task limited to one area of a company. Although product architecture may seem to be primarily an engineering decision, in reality, product architecture choices have impacts across the company. architecture affects product, process, and organization. Product Thus, many segments of a company should be involved in the definition of a product architecture. Firstly, product planners and marketing managers must decide which markets to enter, what market segments to target, and which customer needs to address.34 In doing so, they must understand the long-term development strategy of the company, and select markets and targets based on this strategy. Secondly, system engineers must decide what product architecture to use to effectively share components, processes, and knowledge. To do this, system engineers must understand the long-term development strategy to ensure that the selected architecture will be useful for both current and future products. The most effective product architectures usually arise when distinct disciplines in an organization, including marketing, management, design, and manufacturing, work together to define an architecture. However, often such groups are unaccustomed to working together and sometimes interact very little. In short, determining an effective product architecture needs to be a company-wide, collaborative initiative; product architectures derived by an isolated group within the company, stand little chance of being effective for the company as a whole. Even if the organizational barriers to product architecture are overcome, the job of determining an actual product platform is still very difficult. This task is often left to senior systems engineers who, through time and experience, have gained a solid overview of the product and thoroughly understand the many product goals. What makes this task difficult is that there are many different approaches to determining product architecture, with the spectrum ranging from engineering-based approaches to operations-based approaches. Engineering-based approaches tend to focus on functional issues such as manufacturing and assembly, and strive to create product platforms so as to optimize these areas of the product development process. Operationsbased approaches to product architecture tend to focus on non-functional issues, including market variety, serviceability, and supply chain issues. Developing successful product architectures usually involves keeping both functional and non-functional goals in mind and making trade-offs between these often-competing objectives. Given the nature of this multi-attribute optimization, product architecture decisions are often best made by senior systems engineers. 14 1.3 Scope of this Thesis There are clearly numerous issues to be addressed in determining and implementing a successful product architecture. This thesis focuses on the task of actually defining a product architecture. The work presented here takes an operations-based approach to the problem, addressing the influence of certain non-functional issues. Specifically, this thesis addresses how serviceability, technology change rate, and supply chain influence product architecture. Detailed methodologies to incorporate such concerns into product architecture decisions are developed and presented. These methodologies are intended to be used by product developers, both experienced and inexperienced, in determining effective product architectures. 1.4 Organization of this Thesis Chapter 2 defines some of the terminology relevant to product architecture. This chapter also explores previous work in the field, providing the background for the work presented here. Chapter 3 describes the importance of functional modeling and presents function structures as a tool for making such models. Flow-based modularity heuristics, used to functionally determine product architectures, are also presented. Chapter 4 introduces a methodology to incorporate serviceability into product architecture decisions. Equations to calculate serviceability costs for different modularity schemes are presented and demonstrated on an actual system. Chapter 5 examines the influence of technology change rate on product architecture decisions. Issues that should be considered when determining technology change rates are discussed. A methodology to use these technology change rates in determining product architecture is presented and demonstrated. Chapter 6 presents a methodology to incorporate supply chain concerns into product architecture decisions. Supply chain issues are explored, focusing on how such issues can influence product architecture. The methodology is then demonstrated on an actual system. Chapter 7 summarizes the work presented in this thesis. Methods to determine an overall product architecture are discussed and recommendations for future work are made. 15 16 2 Product Architecture As shown in Chapter 1, selecting the appropriate product architecture can have serious ramifications for product and company success. Because of this, research in this field, and in product development as a whole, has experienced significant growth in recent years. This chapter defines some of the terminology used in product architecture. It also provides a brief overview of recent research in this field, thereby laying the foundation upon which this thesis is built. 2.1 Terminology Product architecture is the scheme by which the function of a product is allocated to physical components.' It is concerned with the arrangement of functional elements, and with how those functional elements map to the physical components of the product. Product architectures are divided into two main types, integral and modular. Integralproduct architectures commonly have multiple functions satisfied by a single component. Modular product architectures,on the other hand, typically exhibit a one-to-one mapping between functions and components. Integral and modular architectures each have their own advantages and disadvantages. Products with an integral architecture have components that perform numerous functions. This technique, known as function sharing, often allows integral designs to be better optimized than modular designs.2 For example, consider a typical carpenter's hammer.3 The head of such a hammer has an integral architecture. A single cast piece features the flat, head end, which can be used to drive nails, and the pointed, claw end, which can be used to remove nails. This function sharing allows for an elegant and efficient solution. In general, arguments for integral product architectures stem from technical and performance-based needs.4 Products with a modular architecture are typically comprised of a set of product modules, each of which only performs a single or a few functions. Product modules are thus defined to be sub-systems within a product that are bundled as a unit and that serve identifiable functions. Products composed of such product modules are generally easier to change, as modules within the product can be replaced individually, instead of replacing the entire product. This has important implications for product variety, time-to-market, product upgrade, and serviceability, among others. For example, consider a typical home stereo, where, due to standardized interfaces, the receiver, tape player, CD player, and speakers can be mixed and matched from 17 various different manufacturers.5 Therefore, if a single component in the system needs to be changed, the entire system is not disrupted. This ability to easily undergo change must be weighed against the disadvantages of a modular architecture, which include lack of product differentiation and reduced product performance. In general, the arguments for modular product architectures, and thus product modules, tend to be based on business concerns. A product platform is the set of assets that are shared between products. include components, processes, knowledge, people, and relationships. These assets can Such sharing across products usually requires the use of an architecture that is somewhat modular in design. As discussed in Chapter 1, product platforms can have considerable advantages, including shortened development time, lowered development cost, reduced product risk, and increased market variety. The set of products based on a product platform constitutes a productfamily. 2.2 Managing Product Architectures Much literature has been written addressing the organizational difficulties companies face in planning for product architectures. Wheelwright and Clark suggest the use of an "aggregate project plan," which helps a company to manage the set and mix of projects under development.' In an "aggregate project plan," projects are categorized as one of five types: breakthrough, platform, derivative, research and development, and partnered. Once the complete set of current and future product offerings is well categorized, a company can better select useful product platforms. Wheelwright and Clark stress the importance of platforms and believe that such projects should constitute a central part of the "aggregate project plan." The importance of involving engineering, marketing, manufacturing, and senior management in the up-front planning of product platforms is also stressed. Robertson and Ulrich present a methodology to help define a product architecture by analyzing the distinctiveness and commonality in a design.' 0 They propose a process that relies on three information management tools: a product plan, a differentiation plan, and a commonality plan. Product developers can use these three tools to create a product architecture through an iterative process. Robertson and Ulrich also stress the importance of having top management involved in the platform planning process. 18 Meyer and Lehnerd present extensive case studies on platforms, demonstrating both their advantages and challenges." They also suggest methods to help chart the evolution of platforms and derivative products, thereby helping to plan product families. Pedersen examines some of the advantages and disadvantages of platform-based design, and discusses some of the organizational challenges faced by companies attempting platform-based product development. 12 Pedersen comments that implementing a product development approach based on product platforms requires a long-term perspective, both in terms of investment and organizational changes. Pulkkinen, Lehtonen, and Riitahuhta introduce the idea of design for configuration, a method that can be used in the development of configurable product families.' 3 Design for configuration helps enable mass customization. Martin and Ishii present methods that use indices to help quantify the costs associated with providing product variety.14 These three indices, a commonality index, differentiation index, and setup index, help designers to develop products that incur minimum variety costs. Kota and Sethuraman also use an index-based method, employing a Product Line Complexity Index to capture the level of part commonality in a product family." This index incorporates issues including materials, manufacturing, and assembly into an overall score that can be used to evaluate a family of products. 2.3 Determining Product Architectures A great deal of work has also focused on how actual product platforms are determined. Various methods have been developed to identify useful modules from a functional description of a product. A method proposed by Stone, Wood, and Crawford, begins by functionally modeling the product, then applying a set of heuristics to identify possible product modules. is discussed in greater detail in Chapter 3. 16 This method Siddique and Rosen propose a graph grammar approach, which allows product developers to design new product families or increase the commonality of existing families." Product functions are represented using graphs, while relationships between functions are incorporated through the use of grammar rules. These rules can identify core functions, which become the platform, and optional functions, which become unique modules. While functionally based methods can indicate product modules, they represent only one of many possible ways in which modularity can be determined. A great deal of other information can be 19 considered when making modularity decisions. Rechtin and Maier present a set of architecting heuristics and checks that system engineers should consider when creating system modules.18 Ericsson and Erixon introduce many different "module drivers" that should be considered when developing a product architecture.' 9 They also present case studies in which these considerations are used in practice. Moore, Louviere, and Verma use a customer needs-based approach to design product platforms.20 Data from conjoint analyses is used to evaluate product features and drive product platform design. Product features that generate strong returns, and can be produced at a reasonable price, are included as part of the product platform. Product features that are only beneficial to a specific market segment are not included as part of the platform, and are instead add-ons to the platform. Gershenson, Prasad, and Allamneni take a lifecycle view of modularity, looking at how a product can be structured to address issues such as manufacturing, assembly, and service.2 ' Coulter, McIntosh, Bras, and Rosen also examine the role of lifecycle considerations in product architecture decisions. Specifically, they examine how modularity can be used to improve product recyclability. The research described above shows some of the many methods for managing and determining product architectures. This thesis builds on this body of knowledge, introducing new methodologies for defining product architectures. 20 3 Functional Modeling Functional modeling is an important step in the product development process. It allows design teams to completely describe a product by identifying the functions and sub-functions that the product must complete. In a sense, a functional model provides a blueprint for the product.' Such a model, when correctly constructed, should consist of a set of form-independent, discrete, indivisible functions. Each function in the functional model contributes in some way to the overall function of the product. In recent years, additional emphasis has been placed on incorporating customer needs into product development decisions. Functional modeling provides an effective means to do this by allowing customer needs to be translated into functional descriptions of the product. Mapping customer needs to function, as opposed to mapping needs to form, also allows design teams a broader design space in which to work. No longer constrained to the form of a previous solution, product developers are now free to explore blue-sky ideas. Extracting problems out to the functional level in effect broadens the problem and thus the solution set. By removing form and component dependence, designers can gain a fresh perspective on the exact issues to be solved. Functional modeling also plays a role in how a product, process, and design team are organized. Understanding how the overall product function can be decomposed into sub-functions can help in determining how the larger design project should be subdivided, both from a product and process standpoint. It also allows a design team to break down a complex problem into simpler sub-problems. Furthermore, interactions between sub-functions often indicate how much communication is necessary between sub-function design teams. Clearly, functional modeling plays an important role in how products are developed. This chapter focuses on the importance of functional modeling in product development. It explains a brief history of functional modeling and introduces different methods to functionally represent products. Function structures, a specific method to functionally model products, are explained in detail. Heuristic methods for determining product modules from a function structure are then introduced and examined. 21 3.1 Background The idea of functional modeling is not new. In the 1940's, with the development of value engineering, Miles emphasized the importance of functional modeling as the cornerstone for this technique. 2 Of the three fundamental steps involved in value engineering, identifying the functions of a product is the first.3 The idea behind value engineering is cost reduction from the standpoint of function. Therefore, instead of striving to directly reduce the cost of the product, the product is instead decomposed into a functional representation. Each function is then examined to find the least expensive way to perform that function. In functionally decomposing a product, product developers attempt to separate the function that needs to be performed from the actual solution. As part of the value engineering methodology, function analysis system technique (FAST) was introduced.4 This technique features the use of FAST diagrams, which display product functions in a logical sequence and allow the functions to be prioritized. Developed in 1965 by Bytheway, this hierarchical approach categorizes functions based on importance, identifying overall product functions and differentiating them from secondary product functions. Overall, FAST diagrams serve to help identify basic functions, systemize these functions, and provide a creative tool for product developers.7 Functional modeling also has a rich history in German design. In many German product design methods, emphasis is placed on identifying the key functions that a design needs to address. This functional approach is based upon early works by Bischoff, Hansen, Rodenacker, and Roth. From this functionally oriented basis grew the idea of function structures, as described by Pahl and Beitz. 9 A function structure is a type of functional model that consists of a network of sub-functions interconnected by flows. While providing a functional decomposition of the product, it also preserves the relationships between sub-functions, as represented by material, energy, and information flows. This function structure method of functional modeling is used extensively in this thesis, and is described in greater detail in the following section. 3.2 Function Structures Function structures are created as a means to represent products functionally. Establishing an accurate, customer-based representation of product function is an important early step in product development. The product function is the overall intended function of the product, or what the 22 product is to do.' This overall product function is often comprised of numerous sub-functions, which are simply components of the product function. These sub-functions correspond to subtasks that the product must complete. Product functions and product sub-functions can represent the same functional product; the only difference is the level of abstraction. To move from a product function to sub-functions, product developers must ask "how?" To move up the hierarchy, from sub-functions to product functions, product developers must ask "why?" Both functions and sub-functions are typically represented by a verb-object pair, such as "turn screw" or "generate light." Functions represent what the product must do to satisfy the needs of the customer. However, there exist other non-functional constraints to designs, such as cost, size, and reliability. While these are not explicit functions that the product must complete, they are important aspects of the product development process. Such non-functional criteria constitute constraints. Constraints must be satisfied by the product, and thus require consideration during the product development process." Function structures can be developed by utilizing Otto and Wood's method of tracing flows. 2 This flow is then traced through the product, For each customer need, a flow is identified. through a sequence of sub-functions that change the flow. The point of view of the product is always preserved. Thus, items such as hands pass through the product, not vice versa. The sub-functions related to each of these independent flows are then merged into a complete function structure which is comprehensive of the customer needs. Interactions between functions in a function structure are represented by flows. These flows are divided into three types: material, energy, and information. Material flows are concerned with the movement of matter, such as gases, liquids, and solids. Energy flows focus on energy in all forms, including electrical, kinetic, and magnetic energy. Information flows have to do with the transfer of signals. While such signals are often transferred electrically, there is a distinct difference between the flow of electricity as energy and the flow of electricity as information. Functions and flows taken together provide a comprehensive functional view of a product. Each sub-function helps to decompose the overall product function into simpler and smaller units. The flows interconnecting these units show the relationship between sub-functions. They also provide an idea about the complexity of the interactions between various sub-functions within the system. 23 In this thesis, specific symbols are used in representing function structures. Functions are shown as boxes. Flows into and out of functions are represented by three types of arrows: unfilled arrows represent material flows, filled arrows represent energy flows, and dashed arrows represent information flows. These standards are shown in Figure 3.1. Material Energy Information N Function e Material Energy Information Figure 3.1: Generic function with flows. As an example, a function structure for a Black and Decker cordless screwdriver was created. The product is shown in Figure 3.2 while the function structure is shown in Figure 3.3. The function structure shows various distinct product sub-functions, each in its own box. Arrows between the boxes show the various flows. The system boundaries, where the user interacts with the product, exist where these flows enter and exit the system. Note that only product functions are shown; functions completed by humans and other systems in the process of inserting or removing screws are not shown. Also, only the objects the device actually interfaces with are part of the function structure. Objects outside this system are shown as flows into and out of the structure. Figure 3.2: Black and Decker cordless screwdriver. 24 Electricity Store Electricity Noise Heat' Electricity C0r to Motion Transmit Electricity Inputlehi Switch Power Thumb Signal Thumb Screw Transform NTs(-- Hand L Hand Hot Turning Screw Turn Screw Transmit Power Rotation Permit Bit Positioning Bit Position Force . -Reaction Hand Hand Force Bit Register Bit Un-lock Bit Secure Bit Release Bit Bit Force into opposite hand Bit v Secure Figure 3.3: Function structure for a Black and Decker cordless screwdriver. The gray area shows the system boundaries. Function structures can be used at different stages of the product development process. They are sometimes used in very preliminary research and development efforts to help conceive and develop new technologies that are not yet physically viable. In this use, the ability of function structures to represent the product in a form-independent manner allows designers to generate new ideas and conceive alternative product forms. For example, with the Black & Decker cordless screwdriver, which utilizes a rechargeable battery, one could generate a new technology concept that uses a different physical principle, such as compressed air, to power the device. In this instance, functional modeling could be used to describe requirements at a level of detail generic to both physical principles. This use of functional modeling is different from how function structures are used in this thesis. Function structures can also be used later in the product development process, such as for 3 deriving physics, establishing specifications, and specifying interface requirements. 25 For example, function structures can be used to focus on the deployment of a technology into product lines. Once a technology concept has been selected and developed to the point of feasibility, function structures can help determine how to implement this technology concept into actual products. In this thesis, function structures are used to help partition a product concept. This is a problem for which functional modeling, and function structures in particular, is ideally suited. 3.3 Modularity Heuristics Once a functional model of a product is created, it can be used to help in determining product architecture. A set of heuristics, developed by Stone, Wood, and Crawford, can be used to 14 identify useful product modules from a well-developed function structure. This clustering of sub-functions into modules relies on examining flows in the function structure. By grouping subfunctions based on flow heuristics, functional relationships can be preserved. This method relies on three heuristics to derive a modular product architecture: dominant flow, branching flows, and conversion-transmission The dominant flow heuristic examines flows through a function structure, following flows until they either exit from the system or are transformed into another flow. The sub-functions through which a flow can be traced, define a module. More succinctly, a set of sub-functions through which a flow passes, from initial entry or formation of the flow in the system, through final exit or conversion of the flow within the system, define a module. =4 Function FunctIOW --- Function Figure 3.4: Example of the dominant flow heuristic. The branching flow heuristic examines flows that branch into or converge from parallel function chains. Each branch of a flow can become a module. Each of these modules interfaces with the product through the point at which the flow branches or converges. 26 Figure 3.5: Example of the branching flow heuristic. The conversion-transmission module examines flows that are converted from one type of flow to another. A conversion-transmission module converts an energy or material into another form, then transmits that new form of energy or material. In many instances, this conversiontransmission module is already housed as a module, as in the case of an electric motor. Convert Convert Transmit Flow Flow Figure 3.6: Example of the conversion-transmission heuristic. While following these three heuristics results in product modules that maintain physical consistency, they also provide the maximal set of functions to incorporate into a module. There are reasons, however, to further refine these suggested modules into smaller groupings. For example, some of the functions within a maximal module might have excessive lifecycle costs, such as for repair. Similarly, a supplier might be able to provide some, but not all, of the functions within a suggested maximal module. In both cases, the maximal module should perhaps be split. The next three chapters present methods to incorporate such non-functional issues into product architecture decisions. Specifically, the effect of serviceability, technology change rate, and supply chain on product architecture is examined. 27 28 4 Serviceability In many industries, service is a major consideration for product development teams. A great number of high-end products, including photocopiers, automobiles, and airplanes, require regular service and maintenance for optimal performance. Such service concerns must be taken into consideration during the product development process to ensure that products are designed for easy and straightforward upkeep. Ignoring such factors can lead to products that are both difficult and costly to maintain. At the Xerox Corporation, serviceability is a major factor to consider when determining product architecture. Most Xerox products span a broad range of service levels. At the most basic level, Xerox products must be regularly serviced by the customer. Consumable parts, such as paper and toner, must be routinely replaced by the customer. While this may not be thought of as service, it is the most commonly needed type of repair. On a slightly higher level, Xerox products must also be serviced after simple, random, unplanned malfunctions, such as paper jams. Such problems are once again most commonly repaired by the customer. At the other end of the service spectrum are repairs that warrant trained Xerox service agents. These repairs can range from fixing unexpected failures to replacing components as part of a regularly scheduled maintenance plan. To handle such repairs, Xerox enters into Field Service Maintenance Agreements with their customers based on a monthly and per print service fee. Thus, such repairs, and the network of service agents associated with these repairs, represent both a substantial income source and cost to the company. Clearly, with a family of products that require diligent upkeep for optimal performance, serviceability issues play a major role in product design deliberations. In selecting a suitable product architecture, engineers must be well aware of serviceability concerns. In this chapter, a methodology for incorporating serviceability issues into product architecture decisions is presented. This method is then demonstrated on a document system from the Xerox Corporation. 4.1 Related Work Design for serviceability has been receiving additional attention in recent years. Since it plays a role both in the product's lifecycle service costs and in long-term customer satisfaction, this additional emphasis is certainly warranted. Techniques have been developed to help product developers address serviceability requirements during early stages of the design process. One 29 such technique that has gained some popularity is that of Failure Modes and Effects Analysis (FMEA). FMEA, initially developed in the 1960's for the aerospace and defense industries, is a method for addressing reliability issues in the early stages of a design.' By providing information early in the process, designs can be improved to increase product quality, improve reliability, and reduce lifecycle service costs. FMEA helps design teams to identify, evaluate, and eliminate known or potential failure modes of a product.2 FMEA focuses on the occurrence, severity, and detection of failures, assigning risk priority estimates to various failure modes. These risk priority numbers help to organize the various failure modes, prioritizing them to determine which should be addressed first. While FMEA provides a useful tool for product developers, other more probabilistic methods can be used when additional data is available. statistical data can be used. If historical failure rate data is available, such In design for serviceability, it is critical that failure rates are accurately predicted or measured, as these failure rates can play an important role in design deliberations. Because data is available, the examples presented in this chapter use a data driven approach to design for serviceability. However, in the absence of such data, FMEA estimates could have been equivalently used. Other research involving lifecycle engineering has also examined the role of serviceability in product design. Gershenson and Ishii were among the first to address serviceability in design.3 They recognized the drivers of service cost, including part cost, labor cost, and failure rate. This information was incorporated into worksheets that could be applied to analyze any design. Parts of their work, including the basis for serviceability cost equations, directly apply to the work presented here. Modified forms of their equations are used here to evaluate candidate modules and architectures. Ishii also looks at an integrated lifecycle design methodology that would allow engineers to better understand the lifecycle implications of certain design decisions. 4 Ishii presents specific evaluation methods for calculating serviceability costs. While these works focus on the role of serviceability in product design, the work presented in this chapter looks at the role of serviceability in product architecture, a related, but distinct area of research. Work has also been done to look at modularity and its effect on design for the lifecycle. One such method looks at modules from the viewpoint of material recycling, service, and post-life 30 intent. This method relies on measuring modularity based on module correspondence between several viewpoints and coupling between modules. An architecture decomposition algorithm is then applied to partition a product into modules from each lifecycle perspective. The method presented here isolates a specific lifecycle aspect, that being serviceability. It also takes a more numerical and in-depth approach to selecting a product architecture. 4.2 Approach The method to incorporate serviceability into product architecture involves five basic steps, as shown in Figure 4.1. Create a function structure for the given product. Determine part cost, labor cost, and reliability information for each function. Determine candidate modularizations using flow heuristics. Compute serviceability costs for each candidate modularization. Evaluate alternative architectures. Figure 4.1: General methodology for including serviceability in product architecture decisions. The method begins with the construction of a function structure for the given product. This function structure, as described in detail in Chapter 3, should contain all functions that the product must fulfill, as well as the material, energy, and information flows between the functions. Next, each function should be analyzed to determine part cost, labor cost, and reliability information. To determine such information, the parts used in a design are attributed to particular 31 functions. Part cost for a particular function is thus simply the sum of the part costs for those involved in completing that function. Labor cost for a function can be determined by examining the time it takes to replace the set of parts responsible for carrying out that particular function. By knowing individual reliabilities for each part, reliability for a function as a whole can be calculated. New products that are currently under development may not have part costs, labor times, and failure rates from past data, but these quantities can usually be estimated, typically from previous product generations or from developmental prototypes. With this information, serviceability can be incorporated into the product architecting process. Product architecture is first determined by applying the partitioning heuristics, as discussed in Chapter 3, to the function structure created in step one of this method. Functions in the function structure are grouped by following flow-based product partitioning heuristics. Application of such heuristics allows various candidate modularity schemes to be developed, depending on which flows are considered. As mentioned previously, while these ideas provide modules that maintain physical consistency, they often incorporate the maximal set of functions into a module. Frequently, reasons exist why such suggested modules should be further refined. As such breakouts are not obvious, the next step in this method uses serviceability equations to examine serviceability constraints on product modularity. For any modularity scheme generated, it can then be analyzed by applying serviceability equations. These equations utilize the serviceability estimates from step two of this process. By using part cost, labor cost, and reliability information, the serviceability cost for each modularity scheme can be determined. With serviceability costs determined for the entire product under each candidate modularization, the product development team is left to evaluate the potential architectures. At this point, product developers must be aware of the many trade-offs involved in selecting a suitable product architecture. This five-step approach provides a new systematic methodology to architect a product while considering serviceability issues. 32 4.3 Modeling Serviceability Considering serviceability constraints is important when making product modularity decisions for products that require maintenance and repair. In such situations, product modularity needs to address the trade-offs involved between reliability, part cost, and labor cost. For example, increasing part reliability often increases part cost. However, with increased reliability, fewer repairs are necessary, resulting in lower labor costs. Given the nature of these three metrics, the maintenance cost structure for a given product should be understood prior to making modularity decisions. 4.3.1 Basic Service Cost Equations In evaluating service costs, equations can be used to take into account reliability, part cost, and labor cost. The basic service cost equation for any single product failure, independent of service model, is as follows, (4.1) Cservice = Cfixed labor + Cvariable labor + Cparts Where Cservice Cfixed labor Cvariable labor Cpars In the above equation, Cfixed labor = Service Cost for any single service = Fixed Labor Cost ($) = Variable Labor Cost ($) = Parts Cost ($) ($) refers to the labor cost associated with sending a maintenance representative to the customer site. The variable labor cost, Cvariable labor, refers to the labor cost associated with actually repairing the product once at the customer site. The cost of the replaced components is represented by the value Cpas. To calculate the service cost of a product over a given period, the service cost equation must also include a failure rate. This failure rate can take the form of failures per length of time or failures per number of cycles. Once again, this equation is independent of the service model used. After including this failure rate, the equation appears as follows, Clife = (Cfixed labor + Cvariable labor + Cparts) x Fprojuct Where Clife Cflxed labor Cvariable labor CPals Fproduct = = = = = Service Cost for a given period ($ per period) Fixed Labor Cost ($) Variable Labor Cost ($) Parts Cost ($) Product Failure Rate (failures per period) 33 (4.2) By splitting labor cost into labor time multiplied by labor rate, the above equation takes the following form, Clife = [[(Lfixed labor + Lvariable labor) x Rlabor] + (Cans)] X Frodt Where Clife Lfixed labor Lvariable labor Riabor Cpars Fproduct (4.3) = = = = Service Cost for a given period ($ per period) Fixed Labor Time (hours) Variable Labor Time (hours) Labor Rate ($/hour) = Parts Cost ($) = Product Failure Rate (failures per period) The service cost equations presented here are typically applied to consider only the costs borne by the company that produces the product. In short, the equations are normally used only to provide an internal view of serviceability. The equations do not reflect the cost or hardship experienced by the customer. Therefore, companies must also examine their service policy from the customer's point of view. For example, these equations do not consider the cost of down time. While the company that produces the product may not suffer from product down time, their customers may lose significant revenue due to the time it takes to secure replacement parts and repair the product. satisfaction. Excessive down time may also adversely affect long-term customer Companies should be aware of such factors when structuring their products and service departments. For any single product, Equations 4.1 through 4.3 hold true. However, these equations are not entirely correct when failures can occur in multiple independent modules within a product. For these situations, slightly different service cost equations are necessary, depending on which service model is considered. These different service models are explored in greater detail in the upcoming sections. The appropriate service cost equations for each particular model are also presented. First, however, a short background in reliability analysis is provided. 4.3.2 Basic Reliability Concepts Determining and understanding the failure rate of components and modules is not a trivial task. An entire field of research known as reliability engineering exists to better understand issues including reliability, quality, and failure. Given that reliability is defined as the probability that an item will not fail for a specified period under stated conditions, any reliability analysis will, by default, involve probability.6 A basic understanding of probability is thus necessary prior to 34 delving into any reliability analysis. While the methodology to incorporate serviceability into product architecture is not meant to involve a rigorous reliability analysis, it is important that product developers understand the reliability information that they possess and how it should be used. Thus, a few basic concepts directly relevant to the examples presented below are explained. Probability distributions are commonly used to represent reliability for mechanical and electrical components. The most popular and most commonly used probability distribution in reliability engineering is the exponential distribution.' The exponential distribution is simple to use and is often used to represent the reliability of electronic components. It has a constant failure rate, meaning that a component is equally likely to fail at any time. This means that such a distribution has a memory-less quality to it; the probability that a component will fail during some period of time in the future is independent of its age. Therefore, each component always has the property of being as good as new.9 A good example of such a distribution involves that of an automobile tire failing due to a nail puncture. A tire is equally likely to be punctured by a nail at any time during its service. Its likelihood of failure by nail puncture is thus constant and can be represented using an exponential distribution. Such a distribution implies that failure of a device does not take place through wear and deterioration, but rather through a random or sudden event.'0 With an exponential distribution, reliability figures such as mean time between failures (MTBF) are readily available. Given a constant failure rate of X, the MTBF is the reciprocal of the failure rate, or 1/ X. Another model of reliability is the Weibull or "bathtub" distribution. reliability analyses to represent mechanical components. It is commonly used in The Weibull curve, shown in Figure 4.2, has three distinct regions. The first region, or break-in period, features a high but decreasing failure rate. These failures can be attributed to such factors as manufacturing errors, material defects, and missing parts. The next region of the Weibull curve is commonly referred to as the useful life period, and is signified by small and near-constant failure rates. region, failures occur as the result of random and unexpected events. These include power surges, vibrations, mechanical impacts, temperature fluctuations, and moisture variations." middle region is typically the longest of the regions. In this This As the failure rate in this region is the lowest, it is usually the one claimed to be the failure rate for the product. The last region of the Weibull curve is the wearout period. This region features a rapidly increasing failure rate, as components gradually age and deteriorate. Such deterioration of parts can usually be attributed to factors such as corrosion, embrittlement, fatigue cracking, and diffusion of materials.' 2 35 Break-in : period 1 Wearout period Useful life period 191 tEtW Time Figure 4.2: A typical Weibull curve. The failure rates for the three regions of the Weibull curve can be improved through different methods. The break-in period can be reduced or eliminated by burning-in and debugging components before releasing them to the market. This region can also be addressed by implementing stricter quality control measures. The second region of the Weibull curve can be addressed through improved design of the components. By making designs more robust to their environment, the failure rate in this region can be significantly reduced. The failure rate of the third region can be reduced through the design of more durable components and the selection of more durable materials. This region can also be addressed through a policy of scheduled inspections and preventative maintenance. The useful life period of the Weibull curve is of particular interest. By breaking-in components at the factory, and by later replacing components as they enter the wearout period, components in the field can be limited to those in their useful life region. This region encompasses the period from tE, the time a component is released to the market, to tw, the time a component is replaced. This period is shown in Figure 4.2. In this useful life region, the failure rate is constant and can 3 be modeled using the exponential distribution mentioned earlier.' 36 In the following examples, systems are analyzed using two different service models. These service models, while both valid, lead to different system reliability figures, and thus different serviceability equations. This in turn can lead design teams to different product architecture solutions. The first reliability model discussed is referred to as the Random FailureModel. The second model discussed is referred to as the Periodic Maintenance Model. Which model is actually used depends on each company's approach to service and maintenance. 4.3.3 Random Failure Model The first service model considered incorporates random component failures. In this model, components fail according to probability distributions. For simplicity, a constant failure rate is used in this example, although other failure functions can also be used. If a constant failure rate is applied, the component can be modeled as if it is in the useful life period of the Weibull curve, as seen in Figure 4.2. This constant failure rate has a memory-less quality to it, meaning that its failure rate is independent of age. This means that no matter how recently a component has been replaced, its failure rate remains constant. Once again, the tire and nail example applies. A tire is equally likely to fail due to a nail puncture when new as it is when nearing the point of needing replacement. The assumptions for this situation are important to note. It is again assumed that components are tested and broken in prior to use in the field, thereby avoiding the high failure rates associated with the break-in period of the Weibull curve. The constant failure rate, X, is determined according to the MTBF data, where MTBF equals 1/k. While this failure rate increases as the components enter the wearout period of the Weibull curve, it is assumed that the maintenance point, tw, is set such that most of these components are removed prior to or early on in the wearout period. Thus, the failure rate does not increase substantially, if at all, before a part is removed. Therefore, for the life of a component, the failure rate can be simply modeled as X. The assumptions for this model are representative of many failure modes of consumer products, including photocopiers and automobiles. For example, consider again the case of automobile tires. Assuming that tires are thoroughly tested prior to leaving the factory, the break-in period is avoided. During the tire's life, its chances of random, unexpected failure remains relatively constant. Such failures can be attributed to chance events such as nail punctures. As the tire approaches 40,000 miles, the failure rate begins to increase slightly, reflecting wearout. However, before failure rates increase drastically, automobile owners typically replace the tires, 37 - -- -11M-OW- Mm- -- - - thereby starting the cycle again. The plot of failure rate versus mileage is shown in Figure 4.3. Multiple tire lifecycles are shown. The first set of tires is replaced at twi, while the second set of tires is replaced at tw2. If both of the tire sets have the same failure rate during their useful life periods, then the failure rate is constant over both sets of tires, from time 0 to time tw2. 0.007- Useful life period 1 0.008- / T ____________________T Wearout period I I Useful life period 2 0.005 - Wearout period 2 10.004- U.O.003- 0.002 - 0.001 - tw2 tw1 - 0 0 10000 20000 30000 50000 40000 0000 70000 80000 90000 100000 Mileage (miles) Figure 4.3: Plot of failure rate versus mileage for automobile tires. The use of this method has implications for module reliability. In this case, the reliability for a grouping of independent components is a function of the product of the reliabilities of each of the components that make up the module. Therefore, the overall reliability of the module declines as more components are included. For example, consider a module comprised of three independent components: one with a reliability of 0.98, one with a reliability of 0.95, and one with a reliability of 0.90. If the components have independent failure modes, and assuming that the failure of any component causes failure of the entire module, then the module reliability is the product of these three reliabilities, or 0.84.14 Using the Random Failure Model to analyze service costs leads to specific serviceability equations. For example, consider two modularity schemes, A and B, as shown in Figure 4.4. Modularity scheme A groups both functions into a single module, while modularity scheme B splits the functions into two separate modules. 38 A B Figure 4.4: Two different modularity schemes, A and B. Intearated Modularity Scheme (A) usine the Random Failure Model As discussed previously, service costs are comprised of three main components: fixed labor cost, variable labor cost, and part cost. For this example, service costs are broken into these three categories. For the integrated modularity scheme A, the equations for these costs per period are as follows, Fixed Labor Cost A Cfixae labor = (Lfixed x Riabor) X Fu (4.4) 2 Lfaxe = Fixed Labor Cost ($ per period) = Fixed Labor Time (hours) Riabor = Labor Rate ($/hour) FIU2 = Failure Rate for Union of I and 2 (failures per period) Where Cfix labor Variable Labor Cost A (4.5) Cvariable labor = (Lvariable12 x Riabor) X Fiu2 Where Cvarjable labor Lvariable12 Riabor = Variable Labor Cost ($ per period) = Variable Labor Time for Functions 1 and 2 (hours) = Labor Rate ($/hour) = Failure Rate for Union of 1 and 2 (failures per period) Part Cost A (4.6) Cparts= (P1 + P2) xFIu2 Where Cpar. P P2 FIU2 = Part Cost ($ per period) = Replacement Part Cost for Function 1 ($) = Replacement Part Cost for Function 2 ($) = Failure Rate for Union of 1 and 2 (failures per period) The overall service cost equation for modularity scheme A is thus, CA = [[(Lfixed + Lvariablel2) x Riabor] + (P1 + P 2 )] xFIU2 39 (4.7) Split Modularity Scheme (B) using the Random Failure Model The split modularity scheme B is analyzed in a similar manner, once again separating the service cost into fixed labor cost, variable labor cost, and part cost. The equations for these costs per period are as follows, Fixed Labor Cost B Crixed labor - [(Lfixed x Rabor) x F1] + [(Lfixed x Riabor) x F 2 ] - [(Lfixed x Rabor) x Fn Where Cfixed labor Lfixed Rlabor F1 F2 Fin2 = = = = = = 2] (4.8) Fixed Labor Cost ($ per period) Fixed Labor Time (hours) Labor Rate ($/hour) Failure Rate for Function I (failures per period) Failure Rate for Function 2 (failures per period) Failure Rate for Intersection of 1 and 2 (failures per period) Variable Labor Cost B Cvariable labor = [(Lvariablel Where Cvariable labor Lvariablel Lvariabe2 Riabor F1 F2 x Rlabor) = = = = = = x F1] + [(Lvariabe2 x Rlabor) x F 2] Variable Labor Cost ($ per period) Variable Labor Time for Function 1 Variable Labor Time for Function 2 Labor Rate ($/hour) Failure Rate for Function 1 (failures Failure Rate for Function 2 (failures (4.9) (hours) (hours) per period) per period) Part Cost B Cparts= (P x F1 )+ (P 2 x F 2) Where Cparts P P2 F1 F2 = = = = = (4.10) Part Cost ($ per period) Replacement Part Cost for Function 1 ($) Replacement Part Cost for Function 2 ($) Failure Rate for Function 1 (failures per period) Failure Rate for Function 2 (failures per period) In Equation 4.8, Fin2 represents the simultaneous failure of both Function 1 and Function 2. The last term in Equation 4.8 is included to take into accout the fact that a simultaneous failure, although involving the failure of both Function 1 and Function 2, only requires one service trip. The reason for subtracting the simultaneous failures is perhaps best shown through a Venn Diagram, as seen in Figure 4.5. The failure region is the entire area within the two circles. 40 M1 2 12 Figure 4.5: Venn Diagram showing failure modes of the split modularity scheme (B). From probability mathematics, it is known that given two events, El and E 2 , which are not mutually exclusive, then, P(EIUE 2)= P(E 1) + P(E 2) - P(EinE 2) (4.11) Using this property, Equation 4.8, the fixed labor cost equation for modularity scheme B, can be simplified to the following, Fixed Labor Cost B (4.12) Cfixed labor = (Lfixed x Riabor) x Flu2 Combining Equations 4.9, 4.10, and 4.12 yields the overall service cost equation for modularity scheme B, as follows, (4.13) x Riabor) + (Pi )] x (Fl)] + [[(Lvariable2 x Rlabor) + (P 2 )] x (F 2 )] + [(Lfixed x Riabor) x Flu2 ] CA = [[(Lvariablel This formula can equivalently be written as follows, CA = [[[(Lfixed + Lvariablel) x Riabor] + (P 1 )] x (F1 )] + [[[(Lixed + Lvariable2) x Riabor] + (P 2 )] x (F 2)] - [(Lfixed x Riabor)x (4.14) Fin Equations 4.4 through 4.14 assume that product failures can be attributed to the module that actually failed. Given this, it is assumed that only the failed module is replaced. The equations also reflect the fact that modules are completely replaced when any failure occurs within. That is, Instead, the entire module is replaced; failed the modules are not inspected or diagnosed. constituent parts are not replaced separately. This is a typical practice in the electronics industry. On the other hand, if some modules have service modes in which their constituent parts can be individually replaced, the equations above can be readily modified. By comparing Equations 4.4 and 4.12, it becomes clear that using the Random Failure Model, fixed labor costs do not serve to differentiate between modularity schemes. The split modularity 41 scheme has failure modes that require the same number of service trips as the integrated modularity scheme. Therefore, given two modularization schemes, one with an integrated module and one with multiple, smaller, independent modules, the fixed service costs for the two schemes are identical. The only difference is whether the service agent replaces different sub-modules at different times or the entire monolithic module several times. With the Random Failure Model, fixed service costs are the same. This result has a significant impact on how product modules are determined. For independent failure modes, the Random Failure analysis means that the fixed labor cost is the same regardless of the modularity scheme used. Therefore, for this service model, only part cost, variable labor cost, and failure rate should play a role in determining product architecture based on serviceability. In general, single module schemes, such as modularity scheme A, will result in higher part costs, as a larger module must be replaced with each service trip. However, with the larger module, it will perhaps be easier to service, thereby lowering the variable labor cost. The fact that the fixed labor cost is independent of the modularity scheme used is based partially on the idea that the system reliability is simply the product of the constituent reliabilities for independent events. While this is a well-accepted calculation, it is not always true in practice, as failure modes are frequently not independent. Often, the reliability of larger multi-functional modules is better than the product of the reliabilities of the constituent functions. Research by Barkan and Hinckley suggests that products designed following design for assembly (DFA) rules have fewer overall defects, and thus higher reliability.15 In DFA, various rules are used to improve the design, the most prominent being to minimize part count. In short, minimizing part count generally increases reliability by eliminating or reducing component failure sources. Based on this, it is highly likely that the reliability of a large module will be better than the reliability of multiple smaller modules. This also makes sense from the viewpoint of system interactions. With fewer large modules, as opposed to numerous smaller modules, the number of interfaces between modules is reduced. With fewer interfaces, reliability is likely to improve. 4.3.4 Periodic Maintenance Model The Periodic Maintenance Model is based on a rigorous maintenance program that replaces parts frequently to prevent component failure. In this model, components are only in use during the useful life period of the Weibull curve. This method of avoiding the first and last regions of the Weibull curve means that components are tested and broken in prior to field use, then removed 42 prior to entering the wearout period. However, unlike in the previous model, the failure rate during the useful life period of the Weibull curve is assumed to be negligible. In some types of service environments, this is not an unreasonable assumption. It was discussed earlier that failures during the useful life region occur due to random and unexpected events. This failure rate can be decreased through robust part design. Assuming that this random failure rate is zero simply means that parts are well designed and do not fail unexpectedly. Instead, parts only fail due to wearout or other time-dependent reasons. This method also assumes that parts will definitely fail upon reaching the wearout period. After passing its use limit, tw, as seen in Figure 4.6, the component has a 100% likelihood of failing. Given this, components must be replaced at or prior to tw to avoid failure. Using this method means that components are never replaced due to failure; instead, all maintenance is of the preventative type, completed prior to failure. 1.2- Wearout period Useful life period 0.8 U0 0.0o.6- 0.4- 0.2- 0 TimeW tE Figure 4.6: Cumulative Distribution Function for components based on the Periodic Maintenance Model. The model assumes no random failures and a 100% failure rate at t 4 , the use limit. Such a maintenance plan is only suitable in certain situations. Usually, it involves an active plan of preventative maintenance, replacing components early and often. In these cases, the tw point is often set early, so as to increase confidence that the product will not enter the wearout period. This results in replacing components that may still have some useable life remaining. However, by replacing them early, failure can be avoided. Such methods are often employed in military 43 and aerospace applications, where the cost of failure can be very high. These models can also be applied to other periodically maintained products, including oil filters and toner cartridges. In each of these cases, regular maintenance and service reduce unexpected failure to negligible, near-zero levels. Using this method for component reliability means that for any grouping of independent components, the reliability for that grouping is driven by the least reliable element. For example, consider a module comprised of three independent components: one which fails after 100,000 uses, one which fails after 150,000 uses, and one which fails after 200,000 uses. With the assumptions given above, each of these components, if not replaced on or before their use limit, will definitely fail. Prior to their use limit, their failure rate is zero. The module as a whole will thus fail at 100,000 uses. Replacing the module at 100,000 uses means that only one component drives the maintenance schedule. The other components, with use limits of 150,000 and 200,000, do not affect the regular maintenance schedule. Using the Periodic Maintenance Method to analyze service costs leads to different serviceability equations. For example, consider again two modularity schemes, A and B, as shown in Figure 4.7. Modularity scheme A groups both functions into a single module, while modularity scheme B separates the functions into two separate modules. A B Figure 4.7: Two different modularity schemes, A and B. Integrated Modularity Scheme (A) using the Periodic Maintenance Model Service costs are again broken into three categories: fixed labor cost, variable labor cost, and part cost. For the integrated modularity scheme A, the equations for these costs per period are as follows, 44 Fixed Labor Cost A Cfixed labor Where = (4.15) (Lfixed x Riabor) X Fmax (1,2) = Fixed Labor Cost ($ per period) = Fixed Labor Time (hours) = Labor Rate ($/hour) = Maximum of the Failure Rates for 1 and 2 (failures per period) Cfixed labor Lfixed Riabor Fmax(1,2) Variable Labor Cost A Cvariable labor Where = (4.16) (Lvariable12 x Riabor) x Fmax (1,2) Cvariable labor Lvariablel 2 Riabor Fmax(1,2) = = = = Variable Labor Cost ($ per period) Variable Labor Time for Functions 1 and 2 (hours) Labor Rate ($/hour) Maximum of the Failure Rates for 1 and 2 (failures per period) Part Cost A Cparts= (PI + P 2 ) x (4.17) Fmax(1, 2) Where Cparts P P2 Fmax (1,2) = = = = Part Cost ($ per period) Replacement Part Cost for Function 1 ($) Replacement Part Cost for Function 2 ($) Maximum of the Failure Rates for 1 and 2 (failures per period) The overall service cost equation for modularity scheme A is thus, CA = [[(Lixed + Lvariablel2) x Riabor] + (PI + P 2 )] x Fmax (1,2) (4.18) Split Modularity Scheme (B) using!the Periodic Maintenance Model The split modularity scheme is analyzed in a similar manner, splitting up the service cost into fixed labor cost, variable labor cost, and part cost. For the split modularity scheme B, the equations for these costs per period are as follows, Fixed Labor Cost B Cfixed labor = [(Lfixed x Riabor) x F] + [(Lfixed x Riabor) x F 2] - [(Lfixed x Riabor) x Fin2 ] Where Cfixed labor Lfixed Riabor F1 F2 Fin2 = = = = = = (4.19) Fixed Labor Cost ($ per period) Fixed Labor Time (hours) Labor Rate ($/hour) Failure Rate for Function 1 (failures per period) Failure Rate for Function 2 (failures per period) Failure Rate for Intersection of 1 and 2 (failures per period) 45 Variable Labor Cost B Cvariable labor Where = [(Lvariablel x Riabor) X F 1] + [(Lvariable2 x Riabor) X F 2] Lvariabe2 = Variable Labor Cost ($ per period) = Variable Labor Time for Function 1 (hours) = Variable Labor Time for Function 2 (hours) Riabor = Labor Rate ($/hour) F1 = Failure Rate for Function 1 (failures per period) = Failure Rate for Function 2 (failures per period) Cvariable labor Lvariablel F2 (4.20) Part Cost B Cpars=(Pi x F 1)+(P Where Cpa, P P2 F1 F2 2 x F 2) = = = = = (4.21) Part Cost ($ per period) Replacement Part Cost for Function 1 ($) Replacement Part Cost for Function 2 ($) Failure Rate for Function 1 (failures per period) Failure Rate for Function 2 (failures per period) Once again using the probability mathematics shown in Equation 4.11, the fixed cost equation for modularity scheme B, can be simplified to the following, Fixed Labor Cost B (4.22) Cfixed labor = (Lfixed x Riabor) x Fiu 2 The overall service cost equation for modularity scheme B is thus, CA = [[(Lvariablel x Riabor) + (PI )] x (Fl)] + (4.23) [[(Lvaiable2 X Riabor) + (P 2 )] X (F 2 )] + [(Lfixed x Riabor) X Fiu ] 2 As before, Equations 4.15 through 4.23 assume that product failures can be attributed to the module that actually failed. Thus, it is assumed that only the failed module is replaced. The equations also reflect the fact that modules are completely replaced when any failure occurs within; failed constituent parts are not replaced separately. By comparing Equations 4.15 and 4.22, it is clear that using the Periodic Maintenance Model, fixed labor costs do differentiate between modularity schemes. When using this model, fixed labor costs do depend on the modularity scheme chosen. The use of the Periodic Maintenance Model also has implications for module reliability. Random failure rates under this model are assumed to be zero. However, the service time must be computed to coincide with the use limit, 46 t,,. For a module, the service time is the minimum service time of any single component. Thus, the failure rate for the module as a whole, is the maximum failure rate among the components. In comparing the Random Failure Model to the Periodic Maintenance Model, it can be seen that for split modularity schemes, as shown with scheme B, the overall service cost equations are the same, regardless of the service approach used. Equations 4.13 and 4.23 are identical. However, the service cost equations for the integrated modularity scheme, as shown with scheme A, are different. Equations 4.7 and 4.18 are not identical. Serviceability costs thus depend on both the service approach and modularity scheme used. 4.4 Example: Xerox Document System The ideas presented above were applied to a document system developed by the Xerox Corporation and are demonstrated here. To begin the process, a function structure of the photoreceptor system from a photocopier is created, as shown in Figure 4.8. A drawing detailing some of the physical components of the photoreceptor system can be found in Figure 4.9. After creating the function structure, the modularity heuristics of dominant flow, branching flows, and conversion-transmission, as discussed in Chapter 3, are applied. This results in various system modularization schemes for the entire photoreceptor system. The modules developed are then analyzed using the serviceability equations. As highlighted in Figure 4.8, two specific areas of this photoreceptor system are examined, the charging module and the cleaning module. Different architectures for each of these sub-systems are explored. 47 Energy Information Mixed Toner and Carrier Light Density Information Toner Toner Density Information Mghtu Figure4.8: F cSelectively Chardsed PR Dica Light to duPR Discharged Meangumre Charge on PR PR Density 7 Discharged PR Develop Toner to Measure Toner Heat PR Untused Toner iclenerratee age Phre I Unused Cleani PRToner bla Moving digital blanket with toner Brush Light to Digital(R PR S Moving digital blanket wtout toner Charge on PR igBlanket Charged Used Toner Particles I I s Dirty I: Blade II \ Energy ank- Endgy I I Regulate Electric I Cleaning Directed Energy -- ~~ Module Energy I FR :Tru s I -- Charge ~~Inform"tion ~ Stnd re Old Field zzEnergyI Digital blanket without toner ChargingmodulehtDevelope Hand StOred Toner Particles Create Electric Field EnIg --- --- Charging Module Figure 4.8: Function structure of the photoreceptor system. The charging module and cleaning module are indicated with dashed boxes. Toner and Carrier Charging module Developer Cleaning moduleDgtabant Photoreceptor (PR) Figure 4.9: Xerox digital document system."6 48 Rotate PR Regulated r--- MoeTransport Digital argeon PR arge 0 mPR Used Toner Porll Uncholged Charged PR Masure Ps T ares Toner Particles - Used Ironer Partices Dirty Light I Brush Charged PR Mn PR Disturb Noutrlize Toner --- --- PR - p PR -ATransfer Blade 4.4.1 Charging Module The charging module contains three functions, as shown in Figure 4.10. These functions are grouped as a module due to the dominant flow of energy, which initially enters the system in the "create electric field" function, and exits the system after the "regulate electric field" function. Grouping these three functions according to the dominant flow heuristic allows all the energy creation and modulation activities to be isolated in a single module. Charging Module Energy -- Create Electric Field Energy Directed Energy Direct Electric Fild Regulate Electric Field IRegulated Energy Charge Information Figure 4.10: Functional diagram of the Charging Module. Given this modularization scheme, representative service costs can be calculated using the previous formulas. Table 4.1 shows the inputs and results of the serviceability equations using the Random Failure Model. For this model, Equation 4.7 is used to calculate service cost. Table 4.2 shows the inputs and results of the same serviceability equations, this time using the Periodic Maintenance Model for reliability. For this model, Equation 4.18 is used to calculate service cost. In these analyses, and in all following analyses, a labor rate of $90 per hour is used. Table 4.1: Service Cost Figures for the Charging Module using the Random Failure Model. Fixed Variable Total Cost Part Failure Rate Labor Labor Labor Module Charging Module front arc shield rear arc shield spring charge wire front insulator block rear insulator block filter charge shield charge grid Time (hours) 1 Time (hours) 0.17 spring clip 49 Cost Cost ($) ($) $105.00 $17.04 (failures/million ($/million copies) copies) $ 7,482.27 61.31 Table 4.2: Service Cost Figures for the Charging Module using the Periodic Maintenance Model. Fixed Labor Time (hours) 1 Module Charging Module front arc shield Variable Labor Time (hours) 0.17 Total Labor Cost Part Cost ($) ($) $17.04 $105.00 Cost Failure Rate (failures/million ($/million copies) copies) $ 7,114.93 58.30 rear arc shield spring charge wire front insulator block rear insulator block filter charge shield charge grid spring clip I Total Cost I * 7,114.93 While this modularization scheme follows the dominant flow heuristic, maintenance costs are relatively high. Grouping these three functions into a single module may not be the best architecture from the viewpoint of serviceability. Other possible architectures are now explored. Rather than incorporating all three functions into a single integrated module, an alternative scheme is to maintain functional independence by separating each function into an individual module. In the modularization shown in Figure 4.11, each function remains separate. This ignores the dominant flow heuristic. Energy Create --->Electric Field Regulation Module Direction Module Creation Module energy -- Direct Electric Fild Directd n - iField Regulate Electric Regulated Energy --- Charge Information Figure 4.11: Functional diagram of the Creation, Direction, and Regulation Modules. Calculations for this modularization are shown in Table 4.3. As demonstrated previously, the type of reliability model used does not affect the outcome. The Random Failure Model and Periodic Maintenance Model yield the same result. Their service cost equations are identical, as shown in Equations 4.13 and 4.23. Table 4.3 shows that a lower maintenance cost results with this split modularity scheme. In Table 4.3, the row entitled "Interactions" refers to the last term 50 in Equation 4.14. This term, as discussed previously, takes into account simultaneous failures. Simultaneous failures are clearly rare, and thus do not have a significant impact on service cost. Table 4.3: Service Cost Figures for the Creation, Direction, and Regulation Modules. These results are true for both the Random Failure Model and the Periodic Maintenance Model. Fixed Variable Total Labor Labor Labor Part Failure Rate Cost Cost Cost (failures/million ($/million Time Time copies) copies) ($) ($) (hours) (hours) Module 376.52 $ 3.01 $120.00 $5.09 0.33 1 Creation Module front arc shield rear arc shield spring charge wire front insulator block rear insulator block $ 0.58 0.005 $6.44 $110.00 1 0.22 Direction Module filter charge shield $ 6,442.73 58.30 $5.51 $105.00 1 0.17 Regulation Module charge grid spring clip $ (0.00) -1.76E-10 0 $ 90.00 0 1 Interactions Total Cost $ 6,819.83 The modularization scheme shown in Figure 4.11 has significantly less associated maintenance costs. The first modularization scheme combined relatively unreliable components, namely the charge grid and spring clip, with other more reliable components. This meant that every time one of these unreliable components failed and was replaced, many reliable components were also replaced. By modularizing based on serviceability, highly unreliable components can be separated, and thus replaced separately. It can be argued that in the first modularization scheme, failure of the module does not necessarily mean that the entire module must be replaced. Instead, as discussed before in a general sense, the module could be disassembled, and only the unreliable components within the module replaced, such as the charge grid and spring clip. However, at Xerox, and at many other companies, this is usually only done with high cost parts. Such an option could be explored. Given the results from the previous two modularization schemes, yet a third modularization might be considered. This modularization, shown in Figure 4.12, features two modules, a creation/direction module and a regulation module. As the integrated modularity equations are 51 used to calculate the service cost figures for the creation/direction module, the service cost results differ depending on the service model used for that calculation. Thus, calculations for the modularization shown in Figure 4.12, using the Random Failure Model, appear in Table 4.4. Calculations for this same modularization using the Periodic Maintenance Model are shown in Table 4.5. Once again, the row entitled "Interactions" accounts for simultaneous failures. Regulation Module Creation/Direction Module Create ---4Electric Field Direct Electric Field Energy Energy Drectqd Energy Regulate --- Electric Fild Regulated Energy Charge Information Figure 4.12: Functional diagram of the Creation/Direction and Regulation Modules Table 4.4: Service Cost Figures for the Creation/Direction and Regulation Modules using the Random Failure Model. Module Creation/Direction Module front arc shield rear arc shield spring charge wire front insulator block rear insulator block filter Fixed Variable Total Labor Labor Labor Cost Time Time ($) (hours) (hours) $120.00 0.33 1 charge shield____ Cost Failure Rate (failures/million ($/million copies) copies) ($) $ 396.56 3.02 $11.53 Part Cost Regulation Module charge grid 1 0.17 $105.00 $5.51 interactions 1 0 $ 90.00 0 spring clip _ 52 _____ _______ ____ _________ 58.30 -1.76E-10 Total Cost $ 6,442.73 $ (0.00) $ 6,839.29 Table 4.5: Service Cost Figures for the Creation/Direction and Regulation Modules using the Periodic Maintenance Model. Fixed Variable Total Labor Labor Labor Part Failure Rate Time Time Cost Cost (failures/million copies) ($) ($) (hours) (hours) Module Creation/Direction Module 1 0.33 $120.00 $11.53 3.01 front arc shield rear arc shield spring charge wire front insulator block rear insulator block filter charge shield 58.30 $5.51 0.17 $105.00 1 Regulation Module charge grid spring clip 0 -1.75E-10 $ 90.00 0 1 interactions Total Cost Cost ($/million copies) $ 395.91 $ 6,442.73 I (0.00) $ $ 6,838.63 This modularity scheme partially preserves the dominant flow heuristic, yet isolates unreliable components to meet serviceability requirements. By partitioning the product in this way, a compromise between flow heuristics and serviceability needs can be reached. The service cost associated with this modularity is only slightly higher than that of the creation, direction, and regulation modules. Further, this scheme is better for other factors, such as assembly. The bundling of two functions into a single module reduces the part count for final assembly. It should also be noted that there is little difference in total cost between the Random Failure Model and the Periodic Maintenance Model. Given the high reliabilities of the components in question, the union of the failure rates, FIu2 , used in the Random Failure Model (Equation 4.7), is not significantly different from the highest failure rate, Fmax (1,2), used in the Periodic Maintenance Model (Equation 4.18). For example, Table 4.1, which uses the Random Failure Model, features a failure rate of 61.31 failures per million copies for the Charging Module. This is obtained by using the union of the failure rates of the constituent modules, namely the Creation Module, Direction Module, and Regulation Module. Table 4.2, which uses the Periodic Maintenance Model, features a failure rate of 58.30 failures per million copies for the Charging Module. This is obtained by taking the failure rate of the least reliable component of this module, namely the Regulation Module. The results from these two reliability models are not significantly different, thereby leading to the small cost difference of only $367.34, approximately 5% of the total service cost. Tables 4.4 and 4.5 show an even smaller difference, with a total cost difference of 53 only $0.66. Given these small differences between the two methods, and given the actual nature of the Xerox service and maintenance methods, only the Random Failure Method is examined in the next example. 4.4.2 Cleaning Module Analyzing a second area of the photoreceptor function structure reveals similar results. Figure 4.13 shows a cleaning module from the photoreceptor system. These functions are grouped as a module due to the dominant flow of used toner, which enters through the "disturb toner particles" function and exits through the "transport toner particles" function. Given this modularization scheme, service costs can be calculated. Calculations are completed using the Random Failure Model for reliability, and are shown in Table 4.6. Blade Brush Cleaning Module Uncharged Toner ne"La Photoreceptor N- Toner Disturb Parbclea- Tie M eParticles Toner (o Used Remove Par Particles Used Transport Toner Particles ate os Poo:cpo Variabl ParFixeds bPhotoreceptor Totalr Figure 4.13: Functional diagram of the Cleaning Module. $2355 36 Failure 231 the Random $100 Mdue Model. Module using Cleaning 4.6: Service Cost Figures for the C0.67 TableClenig brushBad Module Cleaning Module brush bearing Fixed Variable Total Labor Time (hours) Labor Time (hours) Labor Cost ($ 1 0.67 Part Cost ($ $150.00 $23.17 Failure Rate (failures/million copies) Cost ($/million copies) 13.66 $ 2,365.50 brush brush bearing brush gear ground spring cleaner blade auger gear auger bearing auger auger pipe Total Cost 54 $ 2,365.50 Grouping these ten components into a single module results in both a high part cost and a high overall service cost. While this modularization follows the heuristic of dominant flow, perhaps by considering serviceability concerns, a better modularity can be determined. Taking serviceability into account, a different modularity scheme becomes attractive. In the modularization shown in Figure 4.14, each function remains separate, as compared to the integrated modularity scheme in Figure 4.13. While this ignores the dominant flow heuristic, the modules that are formed are effective from the viewpoint of product serviceability. Service cost calculations using the Random Failure Model are shown in Table 4.7. Disturb Module Remove Module Brush Blade Uncharged Used Toner IrOnorU, Transport Module ........ -- Used Used ............ Particles DaRe Pane Transport Toner Tonercl * Toe PhotorceptorPhotoreceptor Photoreceptor Dt Figure 4.14: Functional diagram of the Disturb, Remove, and Transport Modules. Table 4.7: Service Cost Figures for the Disturb, Remove, and Transport Modules using the Random Failure Model. Module Disturb Module Fixed Variable Total Labor Time Labor Time Labor Cost (hours) (hours) I 1 Cost Part Cost Failure Rate (failures/million ($) ($) copies) $180.00 $ 13.73 0.36 ($/million copies) 69.74 $ 13.29 $ 2,056.49 0.01 $ 1.85 -4.9209E-12 $ (0.00) brush bearing brush brush bearing brush gear ground spring____ __ _ _ __ _ _ Remove Module 1 0.67 Transport Module 1 1 $150.00 $ 4.74 Model. $180.00 $ 4.70 1 0 $ 90.00 cleaner blade _____ _______ ____ auger gear auger bearing auger auger pipe interactions 0 Total Cost 55 $ 2,128.08 Once again, the split modularity scheme has lower associated service costs. Grouping unreliable components such as the cleaner blade together with much more reliable parts results in higher overall service costs. By separating expensive reliable components, such as those making up the disturb module, from unreliable components, such as the cleaner blade, savings in maintenance costs can be realized. As before, various redesigns could be completed to reduce part cost and improve part reliability. Reducing the part cost associated with the disturb module could make the single module plan more feasible. Improving the cleaner blade reliability could also make the single module plan more attractive. 4.5 Discussion It is important to point out another method to address the issue of labor cost, that being customer repair. Making modules that are simple enough for customers to repair or replace can greatly reduce overall serviceability costs, as the labor rate would be $0 per hour. However, customers may expect service agents to make repairs, and thus be dissatisfied that they must repair their own system. While such a strategy may reduce short-term service costs, it may adversely affect long-term customer satisfaction and customer repurchase intent. On the other hand, some customers will like the freedom and ease of not having to rely on a service agent. The costs and benefits related to the concept of service must be weighted; specific schemes may not be appropriate for all customers and all situations. There are also possible design approaches that should be considered. For example, while fixed service costs are the same for the Random Failure Model, variable service costs can be made different depending upon the modularization. Combining difficult to handle sub-systems into more easily handled modules makes for reduced variable service times and thus lower overall serviceability costs. Part costs can also be reduced to lower overall service cost. When one monolithic module is not effective given the components used in a design concept, perhaps with more inexpensive elements the single module would be successful. Through redesign, the cost of the module may be reduced, and perhaps in such a way as to meet targeted reliability. Serviceability costs should be considered when making modularity decisions. While heuristic methods to determine modularity are useful to examine possible modularity schemes, other 56 criteria, including lifecycle concerns such as service costs, can often have a large impact on the feasibility of modules. Understanding service cost concerns when modularizing products can result in products that can continue to save time and money throughout their product life. The method presented above presents a means of calculating service cost differences between different modularization schemes. This method provides insight into the effect of service costs on product architecture decisions. This is, of course, only one of the many evaluation criteria that must be considered when making modularity decisions. Other evaluation criteria are presented in the following chapters. The overall decision on how to modularize a product depends on many different considerations, not just those of serviceability. 57 58 5 Technology Change Rate In every complex system, different components in the system can evolve at different rates. Technological improvements often occur on different time cycles, such that while one part of a system may be outdated in a couple years, another part may be outdated in a couple months. For example, consider the case of a desktop personal computer. Some of the components in this system change so rapidly, people sometimes joke that the minute one buys a computer, it is already outdated. However, looking more closely at each of the components that comprise this system, it becomes clear that each has a very different technology change rate. While Intel releases a new processor more than once a year, Microsoft releases a new operating system every couple of years. The physical computer tower, while changing aesthetically, is approximately the same today as it was five years ago. The technology change rate of each of these elements is drastically different. While the physical tower changes very slowly, the operating system changes faster, and the microprocessor changes even faster. The architecture of a typical personal computer allows the user to upgrade individual pieces of the system independently from one another. For example, consider the case of computer hard drives. In 1996, hard drives cost approximately $100 per gigabyte (GB). A 5 GB hard-drive, a large drive for the time, would cost a consumer approximately $500. Five years later, in 2001, given new technologies and fierce competition in the hard drive market, it is not uncommon to see hard drives as large as 40 GB selling for as little as $160. This works out to a mere $4 per GB. The technology of computer hard drives is clearly changing fairly rapidly, providing much more value to the customer. Today, if a computer user runs out of file storage space, he can simply upgrade the hard drive, changing to a hard drive with a greater storage capacity. In doing so, the computer user does not need to sacrifice the rest of his system. The processor, motherboard, operating system, computer case, and numerous other components are unaffected by his decision to upgrade storage space. This capability allows customers to keep pace with technological changes, without constantly sacrificing previous investments. The ability to do this should not be underestimated. Numerous products are not capable of being upgraded; if a customer desires a new technology, often a completely new product must be purchased, even though perhaps only one small part of the product is actually improved. 59 Such a scenario also occurs in product development. Much like the customer who would like to simply upgrade one component without sacrificing the rest of his system, product developers would often like to upgrade one part of a product without having to redesign the entire product. In order to do this successfully, product developers must be able to take a long-term view of product development. In designing the current product, product developers must select an architecture that enables them to later upgrade components as technologies evolve. In doing so, they allow new technologies to be more seamlessly integrated into new products. No longer do technological improvements need to signal drastic redesigns. This chapter presents a methodology for incorporating technology change rate into product architecture decisions. After explaining the methodology, examples based on products from Black and Decker and the Xerox Corporation are presented. 5.1 Related Work New technologies clearly play a major, if not dominant, role in driving product change.' Products are constantly evolving, with each new product striving to provide added customer value, usually through the incorporation of new technologies. With this emphasis on continually improving products, much work has been done to characterize the rate at which technologies evolve. The field of technology forecasting tries to predict what new technologies will occur, when they will occur, and how they will affect the marketplace. Considerable work has been done to characterize the technology life cycle. Perhaps the best known is that of the s-curve, in which new technologies appearing in the marketplace follow an "s"-shaped evolution curve. In an s-curve, technological improvements, as seen in the market, start slowly, then pass through a period of rapid improvement, before finally topping out. S-curves serve to accurately characterize the market behavior of many new technologies.2 Other approaches to characterizing technology life cycles have also been presented. Anderson and Tushman developed a model in which technologies begin with a technological discontinuity, where a new scientific advance or technological insight renders existing technologies obsolete.3 The new technology then enters a period of ferment, during which the technology is researched and prepared for market. Following this, designs enter a period of incremental change until the next technological discontinuity occurs. 60 Recently, much work has been done examining how new technologies affect the marketplace. Christensen presents the idea of disruptive technologies, in which new technologies have the ability to completely alter the marketplace and market landscape.! A disruptive technology often begins by supplying products for a separate market niche. Over time, and with improved technologies, that market niche grows to eventually create a new market, often wiping out some existing markets. Such a scenario may occur with the introduction of digital photography. Initially targeted towards a highly technical, gadget-oriented customer, digital photography had limited market appeal. However, with technological improvements and cost reductions in both digital photography and printing technology, the market possibilities are now very encouraging. Digital photography may be the disruptive technology that replaces the market for traditional silver halide film.5 Another use for technology forecasting is in making technology roadmaps. Technology roadmaps provide a means to show how future technologies will be introduced into upcoming product generations. Understanding which technologies will be used in which products is an important part of product plannning. Such a roadmap can also serve to help bridge the gap between technology development and product development.6 Technology roadmaps have been used by Motorola, Philips, Xerox, and many other companies involved in high-technology industries. While technology forecasting is well-researched, the link between technology change and product architecture is less well-defined. Methods exist to help design for product variety, although such methods usually involve variety in the market at one point in time, not over a period of time, as is the case with evolving technologies. Work by Ericsson and Erixon list technology evolution as one of many drivers of product modularity.9 While the concept of technology evolution is described, their work does not present a detailed method for deriving technology change rates and incorporating them into product architecture decisions. 61 5.2 Approach The general methodology to incorporate technology change rates into product architecture decisions is shown in Figure 5.1. Create a function structure for the given product. Determine technology change rate values for each function. Determine candidate modularizations using flow heuristics. Examine technology change rate values for each candidate modularization. Evaluate alternative architectures. Figure 5.1: General methodology for incorporating technology change rate into product architecture decisions. The method begins with a functional analysis of the product. As before, a function structure for the product should first be created. As described in Chapter 3, this function structure should contain all product functions, as well as the material, energy, and information flows between the functions. Once a function structure is created for a product, each individual function should be analyzed to determine its technology change rate. This technology change rate is usually best expressed as a single number that represents the amount of time until a new technology is ready for market. Coming up with these numbers is not a trivial task; it requires careful examination of the technology landscape, both internal and external to the company. By understanding the rate of technology change for each item in the function structure, technology change rate can be incorporated into the product architecting process. 62 Candidate product architectures are first created by applying partitioning heuristics to the product function structure. These flow-based heuristics can identify multiple modularity schemes that maintain physical consistency. However, the modules identified are also those that incorporate the maximal set of functions into a module. These maximal modules are often not ideal, as other factors besides physical consistency play into modularity decisions. Therefore, these modules should be further refined to better address other modularity concerns. The modularity schemes identified in step three of this method should now be examined by looking at technology change rates. With technology change rates for each function, product modules can now be identified by simply grouping common values. Functions with common values would make effective modules from the standpoint of technology change. Now that different product architectures have been created, the product development team is left to evaluate the various options. In doing so, product developers must consider the many trade-offs involved in selecting a suitable product architecture. The five steps presented here represent a methodology for architecting a product while taking technology change rate into account. 5.3 Technology Change Rate Values In developing an architecture that takes into account technology change rate, perhaps the most critical step is accurately determining technology change rate values for each function. This is a difficult task, as many factors play a role in the development of new technologies. The following sections address this issue, providing a series of questions that, when carefully considered, should help product developers to better estimate the rate of technology change. 5.3.1 When Will the Technology Change? In analyzing the technology change rate of a function, product developers must address the primary question of, "When will the technology change?" Although this question may seem quite straightforward, it is in reality a very difficult question to answer. This vaguely posed question must be greatly clarified before it can be of use in making product architecture decisions. As most technologies evolve gradually over time, it is sometimes difficult to determine what constitutes a "change." In recognizing a "change," product developers must attempt to assign a 63 discrete value to a continuous process. In order to do this, product developers must set a threshold value for technological improvement. Any improvements beyond this value constitute a "change." Given this need for a threshold value, where should such a value be set? From a company's standpoint, technology changes are of little use unless they either improve customer satisfaction or reduce internal cost. Companies can use technology to improve customer satisfaction in various ways. Product performance can be improved while leaving the price paid by the customer unchanged, thereby providing the customer with additional value. Companies can also use technology to improve performance while also raising the price to the customer. In this case, companies must be sure that in the eyes of the customer, the improvement in product capability is sufficient to warrant the increased price. 10 New technologies can also be used not to improve product quality, but rather to lower product cost to the customer or lower internal cost to the company. In each of these cases, there is a cost involved in implementing the new technology. technology can be introduced without some investment of resources. No Often times new technologies require changes in design, manufacturing, supply chain, marketing, and other areas. While the methodology presented here strives to reduce the cost associated with such changes, it is still necessary for product developers to understand the overall cost of such changes. By understanding this, and by understanding the benefits of making such a change, product development teams can decide if introducing such new technologies is appropriate. In many cases, while new technologies may exist, they do not provide the company with a real advantage. The cost of implementing such a technology would outweigh possible financial and strategic benefits. Therefore, it is important for product development teams to know what effect a new technology will have on the bottom line of the company. Only those technologies that will have an overall positive effect should be classified as a "change." Once this threshold value of "change" is determined, product developers must determine how far away they are from having such a "change" ready for the market. While technologies can be developed for limited trial use, the concern here is with when technologies will be ready for the market. This means that issues such as production, supply chain, and others must be resolved. Furthermore, the company must be capable of producing the new technology at the cost points and quality levels necessary. The time estimate for producing this level of technological refinement should be used as the value for technology change rate. 64 This primary question of, "When will the technology change?" must be answered in order to establish valid values for technology change rate. For each function, product developers must determine how much time will pass before a new technology becomes ready for widespread market use. Clearly, a great deal of team and individual judgement comes into play in assigning these values. Product development teams must weigh the costs and benefits of new technologies to both the customer and the company. 5.3.2 Secondary Questions Numerous secondary questions can be asked to help product developers determine when technologies will change. As many factors play a role in technology change, it is important for product developers to understand the entire scope of the situation before determining a value. The more accurately these values can be estimated, the more effective the product architecture will be for future generations of the product. To better predict technology change rate, product developers should consider the following questions: = What is the current technology? " What technologies are being developed? = Where do the current and future technologies fit on a technology s-curve? " How fast have major technological improvements occurred in the past? - How is the rate of technological improvements changing? " What is driving the technological improvements? " Is the new technology one that is beneficial to the customer and/or company? What is the current technology? The first question is simply a means to understand the current state of the technology being evaluated. The current technology can usually be thought of as the latest technology available to the company, either internally or through external suppliers. This technology should be developed to the point that it is capable of being offered to the market at a price point that is attractive to customers. In short, the current technology is the latest market-ready technology. In most companies, the current technology will either be the technology currently on the market or a technology that has been developed and refined since the last product release. 65 What technologies are being developed? The technologies being developed can vary widely, with some naturally being closer to market than others. In considering technologies under development, only those technologies that can realistically be utilized by the company should be considered. For example, technologies being developed in-house can certainly be incorporated into a future product. However, technologies being developed by competitors, or perhaps by the government, will not necessarily be available for use. While numerous different organizations may be working to develop a specific technology, not all of the organizations will produce results that can go to market. Government agencies, while developing key technologies, may not release the technology to the public." In the case of competitors, it would not be uncommon for them to keep the technology to themselves to gain a competitive advantage. Given these obstacles, companies should consider only those technologies that they will legitimately be able to utilize in the marketplace. Where do the current andfuture technologiesfit on a technology s-curve? New technologies can also be analyzed based on where they fit on the technology s-curve. As mentioned previously, technology products often appear on the market in an s-curve, meaning that a plot of one product metric versus time yields an "s"-shaped curve. An ideal s-curve is shown in Figure 5.2. In an s-curve, the start of the curve is marked by infrequent improvements in technology. These initial product offerings fall at the low end of the technology spectrum. As more companies begin to enter the marketplace, a period of rapid innovation occurs. Numerous products are launched in an increasingly crowded marketplace. After this period, the s-curve begins to flatten out, as the rate of technological innovation slows and fewer products are released. By this point on the s-curve, typically only a few mature companies remain in the market. 66 1980 1975 1985 1990 1995 2000 2005 Product Release Date Figure 5.2: An ideal s-curve for technological innovations. Each point represents a product offering. Understanding where a current technology fits on the s-curve can help product developers to determine the rate of technology change. If it is determined that the technology in question is at the start of its lifecycle, developers may predict an upcoming period of rapid innovation. If the technology is already in this period of rapid evolution, developers may predict a slowing of innovation. If the technology is at a very advanced stage, the s-curve may be nearing its technological limit. Understanding these three general cases can help product developers to determine the technology change rate of functions. In the case of technology that is in an advanced stage, product developers must be aware of the notion ofjumping s-curves. When an s-curve begins to plateau, technologies often jump to a new s-curve. In this case, a disruptive technology is developed for which a new s-curve is created. Disruptive technologies, as mentioned earlier, can often lead to large changes in the marketplace and may affect more than simply one function. In the case where technological advances lead to disruptive technologies, product developers must analyze what effect this new technology will have on the product as a whole. It is likely that truly disruptive technologies will bring about large changes in the marketplace and necessitate radical redesigns by product development teams. 67 How fast have major technological improvements occurred in the past? Another way to estimate future technology change rate is to look at past innovation. While the rate of technological improvement changes over time, as shown by the s-curve, past rates of change can provide some information regarding future changes. Assuming that a company has been working with some technology for a period of time, it should have a good understanding of how the technology has evolved in the past. Based on this information, future technology change rates may be estimated. How is the rate of technologicalimprovements changing? While past technology change rates can provide some insight into future technology change rates, product developers must also understand how this rate is changing. This technology change rate may be changing due to both internal and external factors. For example, a company may decide that a certain technology represents a key strategic area, and therefore focus a large amount of research and development funds on developing that technology. In this case, such an infusion of resources may lead to an increased rate of technology change. Technology change rates may also change due to external factors, including development by others, or even through changing government regulations. For example, stricter government emissions regulations on automobiles have almost certainly increased the technology change rate for the electric car industry. What is driving the technologicalimprovements? In understanding technology change rate, product developers should understand what is driving the technological improvements. Once again, this can be attributed to both internal and external forces. Internally, technology change may be driven by the company strategy, or perhaps by certain expertise available within the company. External factors, including development by competitors and by the government, can also drive technological improvements. In determining the drivers of technology change rate, it is important to look outside of the specific industry in which the company is operating. Frequently, technologies are driven by specific industries, such as the aerospace and defense industries. Only over time do these technological improvements become distributed across a broader range of industries and applications. Is the new technology one that is beneficial to the customer and/or company? Lastly, it should be verified that the new technology is actually beneficial to either the customer or the company. As discussed previously, technological improvements should only be 68 implemented if the benefits outweigh the costs. This last question simply serves as a check to ensure that the new technology provides real benefits. It addresses the fact that while new technologies may exist, they may be of no use to the customer or company. If the new technology does not benefit the customer or the company, then the technology being considered should not be implemented. Instead, perhaps more development is necessary before such a technology will positively affect the marketplace. This list of questions provides a solid starting point from which to examine technology change rate. While some of the questions may seem redundant, the goal is not to find exact answers for each, but merely to help product developers come up with a single value for rate of technology change. In doing so, this list helps to illuminate some of the information that should be considered. Once a technology change rate value has been determined for each function, product architecture decisions can then be made. It should be noted that while technology change rates are assigned to functions, coming up with these values often requires analyzing specific solutions, not general functions. Product developers must be sure to look outside the scope of current solutions to evaluate technological breakthroughs that may be on the horizon. The possibility always exists that an entirely new technology could replace the existing solutions. Because of this, product developers should be aware of disruptive technologies that may completely alter the market landscape. No matter how rigorously technology change rate values are derived, they are still merely predictions of the future. It is impossible to predict such values with absolute certainty. Therefore, while elaborate methods may be used to try to more accurately determine values for technology change rate, the methodology presented in this chapter should provide useful product architecture suggestions, even with less accurate values. This product architecting tool, while meant to require a careful consideration of technology change rate, is not meant to require rigorous predictive models. Clearly, if such models are available, they should be utilized. However, in the absence of such models, careful deliberations should suffice. A further complication in predicting technology change rates is that such a process is very susceptible to individual biases. Researchers and developers may feel as if new technologies are only months away from viability, when in fact the technologies may be years away. It is common for people to underestimate the amount of time and resources involved in developing new 69 technologies. This is well evidenced in numerous companies involved in product development. It is a rare project that comes in on-time and on-budget; most projects seem to exceed both cost and time predictions. Up to this point, it has been suggested that technology change rate values should be time estimates of when a new technology will become available. For example, functions should be labeled with technology change rate values such as "6 months," "1 year," or "5 years." However, there exists an alternative approach that allows more flexibility in assessing technology change rates. Instead of coming up with time estimates, product developers can instead come up with generation-based estimates. A generation-based estimate would require product developers to predict how many product generations will pass before a new technology is feasible. Functions would thus be labeled with values such as "1 generation" or "3 generations." In many cases, such a generation-based estimate is easier to determine than an exact time estimate. As an example, consider a company that, on average, releases a new generation of products every three years. In looking at technology change rate and its effect on product architecture, product developers do not need to know exactly when a new technology will become available. Instead, they simply need to know if that technology will be available in three years or in six years. This lower-resolution requirement provides product developers with more leeway in predicting values for technology change rate. Clearly, this method is more useful when product development cycles are long. For industries with rapid product development cycles, accurate prediction of technology change rates is still necessary. 5.3.3 Using Technology Change Rate Values Once technology change rates have been established for each function, and flow-based heuristics have been applied to identify modules, the modularity scheme should be examined. If functions in a module all have approximately the same technology change rate, those functions make sense as a module. However, if the functions have drastically different technology change rates, the functions may need to be separated. By splitting these functions into separate modules, the architecture can better accommodate technology upgrades. 70 The governing equations behind such modularity decisions are as follows, (5.1) Tmodue = min (T, T2 , ... Tn) = Technology Change Rate for a Module = Technology Change Rate for a Single Function, i, where i is Contained in the Module = Total Number of Functions in the Module Where Tmoduie Ti n For a modularity scheme such as the one shown in Figure 5.3, the technology change rate for Module A is simply the minimum of the technology change rates of Function 1 and Function 2. Module A Function 2 iFunction 1 ---- Figure 5.3: Module A with two functions, 1 and 2. The technology change rate of Module A is thus, TmoduleA =min Where TmoduleA T, T2 (5.2) (TI, T2 ) = Technology Change Rate for Module A = Technology Change Rate for Function 1 = Technology Change Rate for Function 2 These equations are now demonstrated on examples from Black and Decker and the Xerox Corporation. 5.4 Example: Black and Decker Screwdriver The ideas presented above are first applied to a cordless rechargeable screwdriver developed by Black and Decker. The screwdriver is shown in Figure 5.4. This is the same Black and Decker cordless screwdriver examined in Chapter 3. Figure 5.4: Black and Decker cordless screwdriver. 71 As the first step in this methodology, a function structure of the product is created. This function structure is shown in Figure 5.5. After creating the function structure, the heuristics of dominant flow, branching flow, and conversion-transmission are applied to define candidate modularization schemes. These modules are then analyzed from the viewpoint of technology change rate. The application of this methodology is now shown as it applies to the power module of the Black and Decker screwdriver. This module is highlighted in Figure 5.5. Power Module r ------------------I Electricity Transmit Electricity Electricit Store Electricity 1 I Thumb Noisectri EPower Heat Thumb Signal to Motion Screw qTorque NMse Transform (T,ns Prvent Input Hot Turning Turn Transmit ~~RotationPoeScwSrw Hand STorque Hand Hand Permit Bit Positioning P Bit Position Reaction Force Hand Hand Force Bit Register Bit Bit Secure Bit Bit Un-lock Bit + Bit Secure Bit Release Bit Bit Force into opposite hand Figure 5.5: Function structure of the Black and Decker cordless screwdriver. The power module is indicated with a dashed box. 5.4.1 Power Module The power module, as shown in Figure 5.6, contains two functions. These functions are grouped as a module due to the dominant flow of electricity, which passes through both functions. In grouping these two functions together, all functions that deal exclusively with electricity are isolated as a single module. 72 Power Module Electricity Store Electricity Electricity Transmit Electricity Electricity Figure 5.6: Functional diagram of the Power Module. Given this modularization scheme, it can now be analyzed from the viewpoint of technology change rate. In this example, exact technology change rates are estimated. However, generationbased estimates could have equivalently been used. Each function is individually analyzed to determine its technology change rate. To accurately arrive at the necessary values, the questions presented earlier are carefully considered. The first function in the power module is the "store electricity" function. This function is currently being completed through the use of a Nickel Cadmium rechargeable battery, a technology that has been on the market for quite some time. At this point, Nickel Cadmium battery technology has clearly reached the mature region of its technology s-curve, as few new innovations have been introduced in recent years. technological limit. The technology has definitely neared its Currently, numerous new technologies are being developed. The most market-ready of these new technologies is that of the Lithium Ion rechargeable battery. The Lithium Ion battery is on a different s-curve from that of the Nickel Cadmium battery. While also at an advanced state on its technology s-curve, Lithium Ion technology is still undergoing slight improvements. It has not yet reached its technological limit. In the past, battery technologies have seen a major improvement on the average of every three years. However, with the automobile industry's push to develop electric and hybrid cars, research into new battery technologies has increased considerably. Given this, it would not be unreasonable to predict that the rate of technological improvements will also increase. However, while such innovations are being developed, Black and Decker may not be able to utilize many of these improvements. Some of the most useful of these new battery technologies may be patented or otherwise protected by the automobile companies. Also, the technology for large batteries, such as those used in the automotive industry, may not scale down to smaller batteries, such as those used in Black and Decker power tools. Given this information, product developers may predict that the rate of technological improvement in the battery industry will increase only slightly, perhaps to a figure on the order of two years. 73 The final issue that needs to be addressed involves the usefulness of new technologies to the customer or company. For this product, battery life is a major concern for customers. Extended battery life would certainly be of use to virtually all customers. Such an improvement could help to differentiate Black and Decker in the crowded power tool market. Given this, improving technologies on the time cycle of two years, or faster, is certainly justified. The second function in the power module is the "transmit electricity" function. This function is currently being completed through short lengths of plastic-coated, stranded copper wire and small metal contacts. Such wiring and contacts have been available for decades. While minor improvements may still be possible, this is a technology that has definitely reached a very mature state in its technology s-curve. The technology has probably hit its technological limit. New material technologies may be under development, perhaps in an attempt to improve conductivity or improve insulation. However, while such research is probably taking place, it is most likely not a high priority. In the recent past, wire and contact technology has progressed quite slowly. Few major technical improvements have occurred in the past decade, and there exists little reason to believe that this rate of technological improvement will quicken. Perhaps specific industries, such as the defense industry, may be researching new technologies in this field. However, these technological innovations may be years away from introduction to the public. Perhaps most importantly in the analysis of the "transmit electricity" function, is the question regarding benefit to the customer or company. The current wire and contact technology performs well; it is not an area of consumer complaint. An improvement in how electricity is transmitted would most likely go unnoticed by the customer. From the company's perspective, the current technology is neither expensive nor difficult to produce. Given this, there is little to suggest that this technology should be improved. Taking all of this into account, the technology change rate for this function is quite slow, perhaps on the order of ten years. Now that technology change rates have been established for each function in the module, the modularity scheme should be re-examined. Given the drastically different technology change rates of the power module, it may be best to separate the functions. By splitting up the "store electricity" and "transmit electricity" functions into separate modules, the architecture can better 74 accommodate technology upgrades. The new modularity scheme for the power module is shown in Figure 5.7. Technology change rate values are indicated for each function. 10 years 2 years E ecity Electricit E lectrcity Electrct Electr ity Figure 5.7: New modularity scheme for the Power Module. Technology change rate values are indicated for each function. Interestingly, the latest cordless screwdriver produced by Black and Decker isolates the "store electricity" function as its own module. As part of the Black and Decker VersaPak line of products, the cordless screwdriver, as shown in Figure 5.8, is powered by a removable, rechargeable battery, as shown in Figure 5.9. In the VersaPak line of products, the battery is no longer an integrated part of the system, but instead a module that can be separately attached to, and removed from, the various power tools. One reason why such a product architecture was selected becomes clear upon looking at a function structure with technology change rate values. The function structure shown in Figure 5.10 shows the technology change rate for each function. Clearly, the "store electricity" function has a much faster technology change rate than the rest of the system. By isolating this function, product developers allow the product to be easily updated when new electricity-storage technologies become available. Much like the example with the personal computer, technology enhancements to this line of power tools can now be made without sacrificing the other components in the system. Figure 5.8: Black and Decker VersaPak cordless screwdriver. Figure 5.9: Black and Decker VersaPak removable, rechargeable, battery. 75 r - r ,I 2years 1 - Transmit Electricity Store Electricity Electricity Nose 5 years i ,---- 10years I r------- I r -- ,- Noise, I ConvertI Het-+-Electricity 1nu Siu~lt to Motion -- 5 years---- | ------ - 5 years 5 years Pvent Rtaio Rotation Transform Noise Thumb iga Torqu S Thumbm -- (T, 1 o 10 years Screw ' --- 6 ears | 5 years Transmit Turn Power Screw I Hot Turning Screw Hand Positioning Reaction Force 1 0yars HLn Hand Hand Force ---- --- - - -- - - - - - - - - - Bit Register Biti Bit - - ---- - ears 5years Fyears Bit Secure Bit Un-lock Bit / Bit - - - - - - - - -1 5 years I Bit L Bitcu-re------------ Release Bit -- -- Bit Force into opposite hand Figure 5.10: Function structure of the Black and Decker cordless screwdriver. Technology change rate values are shown for each function. The indicated modules suggest a possible modularity scheme that is effective from the standpoint of technology change rate. 5.5 Example: Xerox Document System The methodology presented above can also be applied to a document system developed by the Xerox Corporation. Once again, to begin the process, a function structure of the photoreceptor system from a photocopier is created, as shown in Figure 5.11. After creating the function structure, modularity heuristics are applied, as discussed in Chapter 3. This results in the creation of various modularization schemes. Each function is then analyzed to determine technology change rates. The candidate modules are then re-examined from the standpoint of technology change rate. In this example, the charging module of the photoreceptor system, as highlighted in Figure 5.11, is studied." 76 Energy Mixed Toner and CarrierInomtnLih PR e Image Generate Heat Toner Density Information Toner Den PnR m ation Measu Toner Den 't Charge DevO NZ Selectively Discharged DToner to PR Light ParR lMeasure on PR D Selectively irect Parea PtCharge Light to PR Charged PR Tor PR Moving digital Unused Toner blanket with toner Ueoe MoigDigtl -d Neutralize Charge on PR Digital toBlanket S Uncharged Used Toner Particles ta EnC Particles R -yPR Used Toner Dirty Pass toMeayurstm Charge to hr Parcles Dal ty e on PR egulated I I Store Old TonerII Charge e. Information I Energy create I Electric DigtalblakI Hand Field Rotate PR Energy E Direct Electric without toner Charge - in-E Regulate ry Hand Charged PR Charged PR Clean PR r ofmthe -yParticles Charged Blade PR Disturb Toner Used Toner irect T Brush PR PR Transfer Toner 5 Moving digital blanket without Tone Uaede Light Stored PartiFcled IL-_-_ Charging .."q'9 ' Module Figure 5.11: Function structure of the pho-1toreceptor system. The charging module is indicated with a dashed box. Crae Enry 5.5.1 Eery O~ve Drct Energyrah Rlaegu lae Dregulated Energ Charging Module The charging module, as shown in Figure 5.12, contains three functions. These functions are grouped as a module due to the dominant flow of energy, which passes through all three functions. Grouping these three functions together allows all the energy creation and modulation activities to be isolated in a single module. Charging Module Create Energy ---- Energy Electric Field Direct Electric Fild Directe Energy Regulate Electric Fild Regulated Energy --- Charge Information Figure 5.12: Functional diagram of the Charging Module. 77 This modularization scheme can now be analyzed from the viewpoint of technology change rate. Given Xerox's longer product development cycles, generation-based estimates of technology change rates are used in this example. While generation-based estimates allow greater flexibility in predictions of technology change rate, actual time estimates could have equivalently been used. As before, each function is individually analyzed to determine a technology change rate value. To accurately arrive at these values, the questions proposed earlier are again considered for each function. The first function in the charging module is the "create electric field" function. The electric field is currently created by passing high voltage through a wire. This method has been refined over multiple product generations and has now reached a relatively mature stage in its development. While this technology has certainly neared the end of its s-curve, small technology improvements can still be made. New technologies are currently being developed that provide a noticeable performance improvement over the existing charging methods. These new technologies are capable of producing a more uniform electric field, which is an important advantage. While these new technologies are also fairly well developed, they are definitely not as well evolved as the current technology. On the technology s-curve, these new technologies have considerable room for future improvements. In the past, new methods for electric field creation have been developed every couple years. Given the importance of this function, a fair amount of internal resources are committed to new research in this area. This current rate of technology improvement will most likely continue unchanged as there is little to suggest that this will be an area of increased attention and resources. As most of the development of this technology occurs in-house, Xerox is the primary driver. Given that Xerox leads the development of such technologies, they should have a reliable idea as to how fast the technology will evolve. With a typical product development cycle at Xerox lasting three years, it is reasonable to predict that a new technology will be available for the next generation product. Therefore, the generation-based technology change rate should be set to one generation. Lastly, it is important to evaluate if such technological improvements will be beneficial to either the customer or the company. The creation of an electric field plays a key role both in image quality and in overall machine speed. As these are two key areas by which a customer evaluates the product, improvements in these areas would certainly benefit the customer. Given this, the 78 pursuit of new innovations in this area, with the goal of incorporating such advancements into the next generation of products, is certainly justified. The second function in the charging module is the "direct electric field" function. This function is responsible for directing the electric field towards the proper location, that being the photoreceptor. Shields and filters are used to complete this function. Directing the electric field is, at least from a technology standpoint, not a particularly challenging problem. Therefore, little research is currently being done to improve this function. The current technology is probably quite advanced on its technology s-curve, although small improvements may still be possible. This technology has not changed noticeably in the past few years. In general, it does not undergo changes as frequently as other parts of the system. Since the role of this function in the xerographic process is unlikely to change significantly, the rate of technology change in the next decade will probably reflect the rate of technology change for the past decade. Given this, it can be estimated that this technology will not change for perhaps two to three product generations. Lastly, the impact of this function on the customer and company should be evaluated. While this function has a small role in image quality and machine speed, it is not the primary limitation on these properties. Therefore, while this function is important, technological improvements in this function will not necessarily result in performance improvements or in increased customer satisfaction. Therefore, there is no real need to improve this function. Given this, the technology change rate should be set to three product generations. Since changing such technologies would not be particularly beneficial to the customer or the company, there exist no real reasons for developing new technologies in this area. The third function in the charging module is the "regulate electric field" function. This function is responsible for ensuring that the proper voltage is passed to the photoreceptor. This is done primarily through the use of a charge grid. New charge grids are always under development, although this is not a particularly active area of research. While the technology is quite advanced, there is still room for improvement. Thus, on the technology s-curve, the current technology, while mature, has not yet reached its technological limit. This technology has steadily evolved over the past few years. New grid designs have been developed to allow a more uniform regulation of the electric field. Technological innovations 79 such as these occur perhaps every few years. Internal research at Xerox has been the primary driver for such innovations, and will most likely continue to be the main driver. However, the rate at which this technology evolves is not likely to change considerably in the upcoming years. Given this information, it can be estimated that the "regulate electric field" function will not significantly improve for perhaps one to two product generations. Lastly, the impact of this function on the customer and company must be evaluated. Regulation of the electric field has a direct impact on the strength and uniformity of the charge applied to the photoreceptor. Thus, it plays a large role in image quality, which in turn plays a large role in customer satisfaction. Therefore, technological improvements in this area are important in the market. Given this, the technology change rate should perhaps be increased, if possible, such that improved technologies can be implemented with every product generation. Now that technology change rates have been established for each function in the charging module, the modularity scheme should be re-examined. As each of these functions have different technology change rates, perhaps they should not be grouped into a module. By separating these three functions, the architecture can now better accommodate technology upgrades. The new Generation-based modularity scheme for the charging module is shown in Figure 5.13. technology change rates are indicated for each function. Energy --- Create Electric Field I generation 3 generations 1 generation Energ Oirectd Direct Electric Field Regulated Regulate -4-*Electric Field Energy -6--- Charge Information Figure 5.13: New modularity scheme for the Charging Module. Generation-based technology change rate values are indicated for each function. 80 5.6 Discussion The method presented above allows technology change rates to be incorporated into product architecture decisions. While the modularity equations associated with technology change rate are quite straightforward, coming up with actual technology change rates for each function is a difficult task. Product developers must carefully consider both their product and the market landscape in coming up with values for technology change rate. Technology change rate, while important, is only one of the many evaluation criteria that must be Other evaluation criteria, such as serviceability considered when making modularity decisions. and supply chain, also affect the selection of a product architecture. Thus, the overall decision on product architecture depends on many different considerations, not just those of technology change rate. 81 82 6 Supply Chain Many of today's products are sufficiently complex that no single company can produce a product alone. Instead, companies are increasingly dependent on suppliers to provide components critical to the product. For example, consider the case of the Ford Motor Company. In the past, Ford has relied on both internal sources and external suppliers to provide the components necessary for Ford automobiles. While some components were provided by suppliers such as Bosch, Johnson Controls, and Goodyear, other components were designed and manufactured in-house. Recently however, Ford has spun off its auto parts division under the name of Visteon. This move comes only a few years after General Motors spun off its own parts division under the name of Delphi Automotive Systems. company. This recent move by Ford shows a notable strategy shift within the Instead of designing and manufacturing parts and components, Ford has shifted its focus to assembly and system integration.' As an assembly and integration house, Ford is now much more reliant on its suppliers to provide the necessary components for its products. Because of this dependence, design of the supply chain at Ford is certainly of great importance. The product architecture of their automobiles is also critical, as product modules must be defined such that suppliers are capable of delivering the functional modules specified. As opposed to choosing a product architecture, then looking for suppliers, supply chain and product architecture should be simultaneously determined. By doing so, product developers can group product functions into product modules that are effective from a supply chain standpoint. With increasing product complexity, and increasing dependence on suppliers, relating product architecture to supply chain is an important capability. This chapter presents a methodology for simultaneously examining supply chain and product architecture. This methodology specifically focuses on how the network of available suppliers can and should influence the product architecture selected. After explaining the methodology, examples based on a Black and Decker product are presented. 6.1 Related Work In recent years, the topics of supply chain design and supply chain management have become increasingly important areas of research. Research in this area has covered a broad range of topics, from delayed differentiation to evaluating the make-buy decision. The wide scope of this field mirrors the complexity of this issue, as many variables factor into supply chain decisions. 83 In his book, Clockspeed, Fine addresses many issues related to the design and management of supply chains.2 Fine stresses the fact that all advantage is temporary; thus, companies must constantly monitor and adjust their product, process, and supply chain to gain or preserve their competitive advantage. The key to long-term success is to avoid hanging on to diminishing advantages. Instead, companies must constantly strive to find the next new advantage. In doing so, Fine suggests using three-dimensional concurrent engineering, in which product architecture, process architecture, and supply chain architecture are simultaneously developed. Fine also investigates the make-buy decision and its importance to long-term company success. The work presented builds upon previous work by Fine and Whitney, in which the make-buy decision process is studied. 3 In stressing the importance of the make-buy decision, Fine presents many cases, including the famous IBM decision in the 1980's.4 When IBM launched its first personal computer, supply chain decisions were made which eventually proved very costly to the company. With a modular product design, the first computer systems assembled by IBM used major components provided by suppliers such as Intel and Microsoft. IBM primarily functioned as the system integrator. However, as time has shown, this was a very costly decision by IBM. Today, consumers purchase computers based on "Intel Inside" and "Windows 2000," not on the fact that IBM assembled it. As it turned out, IBM outsourced what became the most valuable parts of the computer system. Fine suggests that the auto companies, including Ford, must be cautious of a similar fate. The day may come when cars are purchased based on "Bosch Inside," not on the fact that Ford assembled the product.' Much has also been written regarding the power of delayed differentiation, otherwise referred to as postponement. With uncertain demand for products, companies that can keep a product generic for as long as possible in the supply chain, can usually reap considerable benefits. Most of these benefits come from lower inventory costs and an improved ability to meet customer needs. In the literature, cases are again presented showing the value of such postponement. While much research is currently taking place in the area of supply chain design, and much research is going on in the area of product architecture, little has been done to link the two fields. Fine mentions the importance of concurrently designing product, process, and supply chain, yet stops short of presenting methodologies to do so. Ericsson and Erixon also mention supplier availability as one of many drivers of modularity, but do not explore this idea in depth.8 84 Some research in product architecture looks at product modularity from a lifecycle perspective, yet few incorporate supply chain into architecture decisions. Ishii looks into supply chain factors that influence modularity, such as outsourcing strategy and postponed differentiation.9 His methods incorporate supply chain issues by considering how they affect product lead time. Design charts are used to plot product variety versus lead time, thereby allowing product developers to determine inefficiencies in manufacturing and supply chain. The work presented here addresses the supply chain primarily by examining the network of possible suppliers and how that network can drive product architecture. While this addresses the same relationship between supply chain and product architecture that has been explored by others, it provides a new and more detailed approach to studying this interaction. 6.2 Approach The general methodology to incorporate supply chain issues into product architecture decisions is shown in Figure 6.1. Create a function structure for the given product. Determine candidate modularizations using flow heuristics. Examine supplier possibilities for each candidate modularization. Evaluate alternative architectures. Figure 6.1: General methodology for incorporating supply chain issues into product architecture decisions. As in the previous methodologies, the process begins with a functional analysis of the product. A function structure for the product must first be created. As described in detail in Chapter 3, this function structure should contain a network of product functions, interconnected by material, energy, and information flows. 85 Once a function structure is created for a product, candidate product modules can be formed by applying partitioning heuristics. These flow-based heuristics identify multiple modularity schemes that maintain physical consistency. However, the modules identified through these means are also those that incorporate the maximal set of functions into a module. These maximal modules, while maintaining physical consistency, are often not ideal. supplier capabilities, should also play into modularity decisions. Other factors, such as Therefore, the modules identified through the application of flow-based heuristics should be further refined to better address other modularity concerns. The modularity schemes identified in the second step of this method should now be examined by looking at possible suppliers for each suggested module. Certain groupings of functions may fit perfectly with the capabilities of a particular supplier. On the other hand, some modules may present difficulties, as no supplier is currently capable of providing such a functional grouping. In such cases, the module boundaries may need to be changed. Various candidate modularizations should be analyzed by examining which functional groupings most closely fit supplier capabilities. Now that different product architectures have been created and analyzed from the standpoint of the supply chain, the product development team is left to evaluate the various options. In selecting a product architecture, product developers must consider the many trade-offs involved. The steps presented above represent a methodology for architecting a product while taking supply chain issues into account. 6.3 Evaluating Suppliers In developing a product architecture that takes into account the capabilities of suppliers, a criteria for evaluating different suppliers is necessary. However, this is a very difficult task, as many factors play a role in supplier selection. Furthermore, each company, and even each product development team within a company, may value supplier qualities differently. For example, while some may stress component quality considerations, others may be more concerned with cost, while still others may be most concerned with component lead time. Given this, the following section will address some of the many issues that should be taken into consideration when evaluating suppliers. While the list is not comprehensive, it does provide a solid basis on 86 which different suppliers can be judged. How product development teams in turn use this information to determine suppliers, and thus product architecture, is up to each individual team. In order to incorporate supplier capabilities into product architecture decisions, product developers should consider the following list of questions. These questions should be addressed after flow heuristics have been applied to the function structure and candidate modularization schemes have been formed. " Are suppliers capable of delivering the specified modules? " How many suppliers can provide this module? " Are the possible suppliers a good fit with the company? = How do the modules affect the make-buy decision? Are suppliers capable of deliveringthe specified modules? In incorporating supplier capabilities into product architecture decisions, this first question is perhaps the most important. In partitioning a product for supply chain concerns, it is critical that at least one of the suppliers in a company's supplier network is capable of delivering the specified functional module. Situations will sometimes occur in which no suppliers can provide the exact module as specified, particularly when large modules are defined. In such a case, it must be determined whether or not suppliers could be created or grown such that they could later produce the desired module. Clearly, product architectures must be re-evaluated if modules are created that cannot be provided by suppliers. In determining if a supplier can provide a function, it is important to be aware of the quality, time, and cost constraints for the product. There will almost always be suppliers who have the capability to make a specific functional module; however, there will not always be suppliers who can produce such modules given the quality, cost, and time constraints imposed by the product developers. Modules must be available at cost and quality levels that are appropriate given the product under development. The module must also be available on a time frame that is practical for the product given its production schedule. If a module is very complex, so as to make its lead time significantly longer than that of the other components in the system, perhaps the product architecture should be re-evaluated to bring all module lead times within a certain acceptable range. 87 How many suppliers can provide this module? Once it has been determined that the module can be supplied within the given constraints, the overall supplier situation should be evaluated. Product developers should determine how many suppliers are capable of providing such a functional module. For example, perhaps a module can only be produced by a single supplier. In this case, the company must decide if it is willing to be dependent on a single supplier for a particular functional module. While nurturing a strong collaborative relationship with a single supplier can have a positive effect on product development, some would prefer to avoid such a monopolistic situation. In this case, the product architecture may need to be changed in order to create modules which more suppliers can provide. In contrast, it may be determined that multiple suppliers are capable of producing the necessary module. In this case, competition between suppliers may result in higher quality, lower cost, and shorter lead times. In such a situation, the product architecture may be acceptable as is. Given that there are clearly benefits to both the single supplier and the multiple supplier system, it is up to the individual company to determine the right supplier situation for their needs. However, the company should be aware of their desired situation when determining product architecture. Are the possible suppliers a goodfit with the company? Possible suppliers should be evaluated to see if they are a good fit with the company. It is often better, from both a time and cost standpoint, if a company can rely on a few main suppliers for most of its product modules. This helps to keep overall system complexity low. A larger number of suppliers usually means that more time and resources will be spent managing relationships, coordinating logistics, and carefully defining interfaces. Therefore, when a set of suppliers is proposed, the set should be evaluated to see if such a combination of suppliers is efficient and effective for the company. Perhaps a slightly different product architecture could greatly reduce the number of suppliers, and thus reduce the system complexity. How do the modules affect the make-buy decision? The modularity scheme must also be examined from the standpoint of the make-buy decision. Modules should be analyzed to make sure that key technologies and critical knowledge stays within the company. Each function should be evaluated based on its strategic importance to the long-term success of the company. Those functions that are strategically important, or that may become important in the future, should remain in-house. Given this, such functions should not be 88 grouped with functions that are best outsourced. Understanding the strategic implications of modularity decisions will help prevent critical strategic knowledge and capabilities from leaving the company. This list of questions provides a solid starting point to examine product architecture from the viewpoint of supplier capability. Understanding the effect that the supply chain has on product architecture is critical for product developers. While each development team will make their own trade-offs regarding quality, cost, and time, the questions presented above provide a framework by which such issues can be considered. As with most development issues, specific decisions will depend on the product, company, market, and development team. It should be noted that the idea of suppliers extends to internal suppliers, or divisions within a single company that are responsible for providing components and modules to other divisions in the same company. It is important to consider internal suppliers when making product architecture decisions. While most of the questions posed above should still be considered when dealing with internal suppliers, some of the questions, particularly those related to the make-buy decision, are clearly no longer applicable. Also, while the product is examined from a functional perspective, the supplier capabilities are often examined by analyzing specific solutions, not general functions. Product developers must be sure to look outside the scope of current solutions when evaluating possible suppliers. It is likely that with a different physical solution to a particular set of functions, a completely different pool of suppliers becomes available. Product developers must be aware of other possible solutions and suppliers when determining an appropriate product architecture. 6.4 Example: Black and Decker Screwdriver The ideas presented above are applied to a cordless rechargeable screwdriver developed by Black and Decker. The screwdriver is shown in Figure 6.2. This same screwdriver was analyzed previously using the methodology for technology change rate. 89 Figure 6.2: Black and Decker cordless screwdriver. The first step in this methodology involves creating a function structure of the product, as shown in Figure 6.3. The modularity heuristics of dominant flow, branching flow, and conversiontransmission are then applied to the function structure to define candidate modularization schemes. These candidate modules are then analyzed from the viewpoint of supplier capabilities to determine which make sense from a supply chain perspective. The application of this methodology is now shown as it applies to the torque module and power module of the Black and Decker screwdriver. These modules are highlighted in Figure 6.3. Power Module Store Elcticty Electricity Electriciy ElectricitY Transmit El Thumb Noise, IConvert Torque Ice M ISwitch Electricity to Motion Heat *r Iul Thumb Torque ack Transform Noise Input Signal Power Hot Turning Turn Tranit P RRotatione Force Hand Inu Hand HHand oqePermit +Bit Bit Position Hand Force Bit Register Bit E' Bit BitBit Secure Bit / Un-lock Bt Release Bit Bit +~ Bit Secure Force into opposite hand Figure 6.3: Function structure of the Black and Decker cordless screwdriver. The torque module and power module are indicated with dashed boxes. 90 6.4.1 Torque Module The torque module, as shown in Figure 6.4, contains two functions. These functions are grouped as a module due to the conversion-transmission heuristic. In this case, electricity entering the system is converted to mechanical energy, which is then altered and transmitted. In grouping these two functions together, all functions that deal with converting electrical energy to the correct rotational output are isolated as a single module. Torque Module Electricity ConVert Torque Torque ----- Electricity--to Motion heat Figure 6.4: Functional diagram of the Torque Module. This modularization scheme is now analyzed from the viewpoint of supply chain. Specifically, the possible suppliers for such a module should be considered. In considering possible modularization schemes, the questions proposed above are carefully considered. The first function in the Torque Module is typically completed using a small DC motor. The second function transforms the rotational output from the motor, reducing the RPMs and increasing the torque. This is usually done through the use of gear reductions, using sets of gears to step down the rotational speed. There are certainly suppliers that can provide these functions as a single module. However, in many cases, such a module would require either special engineering on the part of the supplier, in terms of both design and manufacturing, or finding a secondary supplier for a sub-system, then integrating their components with those of the secondary supplier. Companies focused on DC motors would need to acquire expertise in gears and transmissions, while companies focused on transmissions would need to acquire expertise in motors. In either situation, substantial investments in time and resources will need to be made on the part of the supplier. While some suppliers may be willing to do this, it could lead to part costs and lead times that are unreasonable for the given product. With no suppliers that are uniquely capable of providing such a module, perhaps other modularity schemes should be analyzed Perhaps splitting this module into a Motor Module and a Transmission Module, as shown in Figure 6.5, could make it easier to find suppliers. DC motors are virtually a commodity, with 91 numerous companies competing in a crowded marketplace. Companies such as Mabuchi Motor, Johnson Electric, and Winston Electric all produce small DC motors for such applications.10 Given that all three of these companies provide ready-made solutions for the function of converting electricity to motion, it may be easier to go with such a supplier. As such motors are already being produced, cost and lead time should be less than for the Torque Module. The Transmission Module can also be easily supplied by a host of gear and transmission suppliers, including SDP-SI, Berg, and PIC." Black and Decker also produces some gears and transmissions for use in other products, and thus has the knowledge and capability to produce such systems. Each of these suppliers can readily provide a wide range of gear and transmission solutions, thereby helping to minimize cost and lead time. By looking at supplier capabilities, perhaps a split modularity scheme is more effective. Motor Module Electricity Convert Transmission Module Torqu Torque ----- Electricity to Motion Iheat Figure 6.5: Functional diagram of the Motor Module and Transmission Module. Grouping the functions into a single Torque Module would allow a single supplier to provide both functions. This may be advantageous in that fewer supplier relationships will need to be managed. Also, with a single supplier for these functions, fewer interfaces will need to be defined. With two suppliers, the interface between the Motor Module and the Transmission Module would also need to be defined. However, the savings realized by having a single supplier for this module may not outweigh the higher part costs and longer lead times. These issues must be carefully evaluated by the product development team. Lastly, the effect of modularity on the make-buy decision should be analyzed. If the Torque Module is used, Black and Decker would be forced to outsource the module. With limited current expertise in DC motor design or production, Black and Decker would be dependent on an outside supplier for these functions. However, as this is not a particularly strategic module, outsourcing may not be a problem. If the split modularity scheme is used, the motor can easily be outsourced to one of the many DC motor producers, including those mentioned previously. Meanwhile, the gear and transmission module can be produced by Black and Decker, using internal design and manufacturing capabilities. By splitting up the Torque Module, commodity 92 components, such as the DC motor, can be outsourced, while other components, such as the transmission, can be kept in-house. As neither component is of particular strategic importance, the effect of modularity on the make-buy decision may not be significant. Evaluating the advantages and disadvantages of these competing modularity schemes, the split modularity scheme appears to be a valid possibility. A split modularity scheme would probably lead to shorter lead times and lower module costs. Also, by splitting the two functions, product development teams can take advantage of external expertise in the DC motor industry and internal capability in gears and transmissions. 6.4.2 Power Module The power module, as shown in Figure 6.6, contains two functions. These functions are grouped as a module due to the dominant flow heuristic. In this case, electricity passes through both functions in the module. In grouping these two functions together, all functions that deal exclusively with electricity are isolated as a single module. Power Module Electricky Electricity Electricity bectricity Elecrict Figure 6.6: Functional diagram of the Power Module. This modularization scheme is now analyzed from the viewpoint of supplier capability. In doing so, possible suppliers for such a module should be considered. As possible modularizations are evaluated, the questions proposed above are again addressed. The first function in the Power Module is completed through the use of a rechargeable battery. The second function simply transmits electricity from the battery to the rest of the system, and is thus typically completed by wires and metal contacts. Battery suppliers are certainly capable of providing the entire Power Module. While their expertise may be in battery technologies, a battery supplier could easily purchase wires and contacts from a secondary supplier and add them to their batteries, as per customer instructions. However, given that their primary focus is on batteries, systems, such as the Power Module, may result in higher prices and longer lead times. Wire and contact suppliers could also supply such a module by purchasing batteries from a 93 secondary supplier and adding them to their components. However, as before, such system designs may result in higher prices and longer lead times. Companies that specialize in individual components may not be interested in putting together systems. Given these considerations, and given cost and time constraints, the number of suppliers for such a module is probably relatively low. Splitting this module into a Battery Module and a Contact Module, as shown in Figure 6.7, could make it easier for suppliers. Numerous companies are capable of producing rechargeable batteries, including Panasonic, Sanyo, and Varta.12 Given that many companies provide readymade solutions for the function of storing electricity, it may be easier to go with such a supplier. As such batteries are produced in quantity, cost and lead time for a Battery Module should be less than for a Power Module. Wire and contacts are basically a commodity, with numerous possible suppliers. Companies including Belden, BICC General, and Alpha are capable of providing the transmit electricity function.13 Given this, acquiring such components from one of these companies could help minimize cost and lead time. Contact Module Battery Module Eletrci Store E Electricit i Transmit Electricity Figure 6.7: Functional diagram of the Battery Module and Contact Module. Grouping the functions into a single Power Module would allow a single supplier to provide both functions. Instead of integrating the two functions in house, Black and Decker could simply rely on a single supplier to provide such a module. This would lessen the number of suppliers, lessen the number of interfaces that must be defined, and lessen the part count in final assembly. However, these advantages must be considered together with the disadvantages of higher part costs and longer lead times. Lastly, the effect of modularity on the make-buy decision should be analyzed. With limited internal expertise in these areas, and a well developed supplier network, Black and Decker could easily outsource such a module. However, the Battery Module may be becoming a more important strategic module, as longer battery life can help to differentiate Black and Decker tools from others on the market. Given this, battery research and development may be a capability that Black and Decker should try to bring in-house. If indeed this is a strategic focus of the company, 94 by splitting up the Power Module, the battery development can be brought in-house while the wiring and contacts can be outsourced. Evaluating the advantages and disadvantages of these competing modularity schemes presents a difficult trade-off. Product developers must carefully examine the issues and, depending on their own evaluation, determine an architecture that fits their supply chain strategy. 6.5 Discussion The method presented above allows supply chain issues to be incorporated into product architecture decisions. The network of suppliers available to a company should be understood by product developers when making product architecture decisions. By doing so, architectures can be determined that fit well with the company's supply chain. While supply chain is important, it is only one of the many evaluation criteria that must be considered when making modularity decisions. Other evaluation criteria, such as those presented in the previous two chapters, also greatly affect the selection of a product architecture. The overall decision on product architecture must take into account all of these considerations, not just those of supply chain. 95 96 7 Conclusions Determining product architecture is a critical step in any product development activity. By selecting an effective product architecture, companies can realize large savings, both during product development and over the life of the product. This thesis shows how non-functional issues such as serviceability, technology change rate, and supply chain can be incorporated into product architecture decisions. The methodologies presented here can be used by product development teams to determine product architectures that are effective for the company. 7.1 Optimization In this thesis, the issues of serviceability, technology change rate, and supply chain are analyzed individually. Each methodology presented looks at only one issue at a time; therefore, each methodology yields a product architecture that is optimized for a single concern. While this is an important step, an actual product architecture must incorporate many different objectives into a single modularity scheme. These objectives include, but are not limited to, serviceability, technology change rate, and supply chain. Therefore, while each specific concern may suggest a different modularity scheme, these often-competing goals must be resolved to yield a single product architecture. The trade-offs involved in making such a decision depend on the overall objectives of the product, company, and development team. Making such a decision can be a difficult task. In defining a single modularity scheme, primary goals for the product architecture must be determined. Once appropriate modularity objectives are selected, different candidate product architectures should be formed. These can be determined using flow-based heuristics, as discussed in Chapter 3, or other partitioning methods. The candidate modularity schemes must then be evaluated to determine how well they meet the overall product architecture goals. At this stage, specific methodologies, such as those presented in this thesis, can be used to help evaluate and adjust these candidate modularizations. Once each candidate modularity scheme has been evaluated from the perspective of each architecting goal, the product development team is left with a multi-attribute decision-making problem. Various methods exist to complete such a multi-attribute evaluation. One of the most basic evaluation methods is that of the Pugh selection matrix. A Pugh matrix can be used to select 97 between multiple different concepts using a set of defined criteria.1 In using a Pugh matrix for architecture selection, candidate architectures are listed in columns, while judgement criteria are listed in rows. One architecture serves as a datum, against which other candidate architectures are compared and rated. For each of the evaluation criteria, architectures under consideration are given one of three ratings, as compared to the datum: better, signified by a plus (+), worse, signified by a minus (-), or the same, signified by an "S". The ratings in each column can then be summed, yielding an overall score for each candidate architecture. While the Pugh selection matrix provides a simple yet effective evaluation technique, more numerical approaches, such as decision analysis, could also be used.2 As an example of this architecture selection process, consider the Black and Decker cordless screwdriver analyzed in the previous chapters. In determining an effective overall product architecture, the primary goals will focus on serviceability, technology change rate, and supply chain. Other goals can certainly be selected, depending on the objectives of the product, company, and product development team. Once these primary goals are established, candidate product architectures should be created. Three possible modularity schemes are formed, as shown in Figures 7.1 through 7.3. Modularity scheme A, shown in Figure 7.1, groups functions following flow-based heuristics from Chapter 3, thereby creating maximally sized modules. Modularity scheme B, shown in Figure 7.2, keeps each function as a separate module, thus resulting in 15 individual modules. Modularity scheme C, shown in Figure 7.3, groups functions following both flow-based heuristics and methodologies presented in this thesis. Thus, the modules formed address the influence of serviceability, technology change rate, and supply chain. 98 Electricity Eletriity Electrici Store Electricity Transmit Electricity I Thumb L Electricity Noise Heat I Convert I Electricity EetiiyPower to I Motion 1 u Thumb Thm Signal Screw Torqub Noise HandHand Ns Transform (T, ~ ~Input Hand Bc Rotation -- Torque I.-ILLL..... Transmit Power Turn ew .P Permit Bit Positioning Hot Turning Screw BtPsto Bit Position Reaction Force Hnd------------Hand Hand Force Bit Register Bit Bit Bitoc Secure Bit BitRel Bit Un-ock Bit Bit Release Bit o, Bit~Secure Bit Force into opposite hand Figure 7.1: Modularity scheme A. This candidate architecture groups functions into maximally sized modules. 99 r ------Store Electriity Electricity r-------i 1 Transmit Electricity Electrici E le I ctric TThumb Transform- Nos I IConver ~ I T,Moi Nde I Roaio n -- oerSrw II -------------- -- ce PlerticitBwitc I r___ to~ostonn II I--- rque Ho Turnin Tur I Trnsi El 1 PnpstTumn Roineaocoiot nFrc r - Hand Hand Force Bt r--,------1 IBi I Register Bit Bit r--------1 IBi'BitI Secure r--,------1 Bt Bit r--,------1 Un-lock Release Bit Bit Bit Secure Force into opposite hand Figure 7.2: Modularity scheme B. This candidate architecture keeps each function as a separate module. 100 r Store Electricity Electricity Transmit Electrici Electricity Thumb Electricity Heat I Convert 4-- Electricity Motion Porto ......- . 1 Thumb T Signa Screw _n_ Transform Noise r Hand I I.I (Tv t to''t(T Rotation Haynd - Turn SrewScrew Transmit co r- --- -- - Permit Bit Torque Hand P Bit Position Reaction Force ~Positioning L Hot Turning Hand Hand Force Register Bit Bit Secure Bit Un-lock Bit Bit Bit Release Bit -, Bit Secure Bit Force into opposite hand Figure 7.3: Modularity scheme C. This candidate architecture groups functions following both flow-based heuristics and methodologies presented in this thesis. Each of these candidate modularization schemes is then evaluated on how well it fulfills the goals for the overall product architecture. As mentioned previously, the primary goals for this architecture will focus on serviceability, technology change rate, and supply chain. Evaluating the candidate architectures on these three metrics involves using the methodologies presented in this thesis. These evaluations are shown in Tables 7.1 through 7.3. Each table lists the product functions from the function structure in the first column. This list of functions is grouped into modules according to the modularity schemes diagrammed in Figures 7.1 through 7.3. Each candidate module is then examined to determine values for serviceability, technology change rate, and supply chain. Serviceability figures for a module represent the cost of service per 100,000 hours, as calculated using the equations presented in Chapter 4. In this example, the Random Failure Model was used. The technology change rate values for a module reflect the fastest technology change rate of the individual functions in the given module. These values were derived using the criteria presented in Chapter 5. The column entitled "supply chain" shows the potential suppliers for each module. Methods presented in Chapter 6 were used to obtain supply chain information. 101 Table 7.1: Evaluation of Modularity scheme A based on serviceability, technology change rate, and supply chain. Serviceability ($1100,000 Technology Change Rate Supply Chain (potential hours) (years) suppliers) 2 Panasonic, Sanyo 10 Internal store electricity transmit electricity $ switch power input signal 67.10 $ 7.10 $ 122.40 5 Johnson, Mabuchi, SDP-SI $ 10.80 10 Internal unlock bit release bit $ 364.80 5 Internal Total $ 572.20 I convert electricity to motion transform (T, o) prevent back rotation transmit power turn screw pert ihandpositioning register bit 102 Table 7.2: Evaluation of Modularity scheme B based on serviceability, technology change rate, and supply chain. Supply Chain Serviceability Technology ($/100,000 Change Rate (potential hours) (years) suppliers) store electricity $ 47.20 2 Panasonic, transmit electricity $ 7.10 10 Belden, Alpha switch power input signal $ $ 14.20 7.10 5 10 Internal Internal convert electricity to motion $ 30.00 5 ouschn, transform (T, co) $ 28.80 5 prevent back rotation transmit power $ $ 28.80 21.60 5 5 Internal Internal turn screw $ 23.10 5 Internal input hand permit bit positioning register bit secure bit un-lock bit release bit $ $ $ $ $ $ 7.20 7.20 100.00 100.00 97.60 97.60 10 10 5 5 5 5 Internal Internal Internal Internal Internal Internal $ 617.50 Total a 103 Be PlC Table 7.3: Evaluation of Modularity scheme C based on serviceability, technology change rate, and supply chain. Serviceability ($/100,000 hours) Technology Change Rate (years) Supply Chain (potential suppliers) store electricity $ 47.20 2 Panaonic, transmit electricity $ switch power input signal 15.60 5 Belden, Alpha $ 7.10 10 Internal convert electricity to motion $ 30.00 5 hnson, transform (T, o) prevent back rotation transmit power $ 88.20 5 SDP-SI, Berg, PIC turn screw input hand $ 7.20 10 Internal permit bit positioning register bit $ 7.20 10 Internal unlock bit release bit $ 364.80 5 Internal Total 567.30 104 Once candidate product architectures are formed and evaluated, a Pugh selection matrix can be used to select a single product architecture. Table 7.4 shows a Pugh matrix used to select a product architecture for the Black and Decker cordless screwdriver. In this selection matrix, modularity scheme A is used as the datum. Modularity schemes B and C are compared to the datum and given one of the three possible ratings. For example, on the criteria of serviceability, Thus, a minus (-) modularity scheme B is compared to the datum, and judged to be worse. appears in the Pugh matrix. On the same criteria, modularity scheme C is judged to be better than the datum. Hence, a plus (+) appears in the Pugh matrix. After each criteria is judged, total The results show that given the three scores can be tabulated for each architecture scheme. criteria of serviceability, technology change rate, and supply chain, modularity scheme C is the best choice. Table 7.4: A Pugh selection matrix for the selection of an effective product architecture given the criteria of serviceability, technology change rate, and supply chain. Candidate Modularity Schemes I- ! 51 7.2 Serviceability Technology Change Rate Supply Chain Total Ii 3- A U U B UME C datum datum + + datum datum S 0 S +2 - + j Future Work This thesis focuses on three important concerns that have serious influences on product architecture. While serviceability, technology change rate, and supply chain are all significant, there exist still more issues that must be considered during product architecture decisions. Issues such as product variety, recycling, and manufacturing can all drive product architecture. Therefore, the three methods presented in this thesis do not constitute a complete set of concerns. Additional methodologies should be developed to address the influence of these other issues on product architecture. By considering more of these influences, it becomes more likely that a product architecture will be effective for the given design objectives. However, taking more influences into account will also make the overall optimization more complex. additional optimization methods should also be explored. 105 Given this, 106 8 Notes Chapter 1 1. http://www.intel.com 2. http://www.honda.com 3. Sanderson, Susan W. and Uzumeri, Mustafa. Managing Product Families. Chicago, Illinois: Irwin, 1997. 4. Stalk, Jr., George and Hout, Thomas M. Competing Against Time: How Time-based Competition is Reshaping Global Markets. New York, New York: The Free Press, 1990. 5. The term "clockspeed," used by Charles H. Fine in his book Clockspeed: Winning Industry Control in the Age of Temporary Advantage, refers to the rate of evolution, usually as it relates to an industry, product, process, or organization. (Reading, Massachusetts: Perseus Books, 1998.) 6. Young, Lewis H. "Product development in Japan: evolution vs. revolution," Electronic Business, June 17, 1991, pp. 75-77. 7. Pine II, B. Joseph. Mass Customization: The New Frontier in Business Competition. Boston, Massachusetts: Harvard Business School Press, 1993. 8. Hounshell, David. From the American System to Mass Production, 1800-1932. Baltimore, Maryland: The Johns Hopkins University Press, 1984. 9. Wheelwright, Steven C. and Clark, Kim B. "Creating Project Plans to Focus Product Development," HarvardBusiness Review, March-April 1992, pp. 70-82. 10. Sanderson and Uzumeri, 1997. 11. Pine, 1993. 12. Sanderson, Susan W. and Uzumeri, Mustafa. "Managing product families: The case of the Sony Walkman," Research Policy, 24, 1995, pp. 761-782. 13. Sanderson and Uzumeri, 1997. 14. Ulrich, Karl. "The role of product architecture in the manufacturing firm," Research Policy, 24, 1995, pp. 419-440. 15. Robertson, David and Ulrich, Karl. "Planning for Product Platforms," Sloan Management Review, Summer 1998, pp. 19-31. 16. Bremner, Richard. "Big, bigger, biggest," Automotive World, June 2000, pp. 36-43. 17. The Audi A4 is produced on a platform called the B5, which is shared with the VW Passat and Audi A6. In 1999, VW sold almost 840,000 cars based on their B5 platform. Bremner, June 2000. 18. Bremner, Richard. "Cutting edge platforms," Automotive World, September 1999, pp. 30-33. 19. Bremner, Richard. "Common knowledge," Automotive World, June 1999, pp. 42-46. 20. Ibid. 107 21. Image taken from Bremner, June 1999. 22. Bremner, September 1999. 23. Bremner, June 1999. 24. Greimel, Hans. "DaimlerChrysler Loses $269 Million," AP Business News, February 26, 2001, pp. 1-4. 25. Ibid. 26. DaimlerChysler owns 33% of the Mitsubishi Motors Corporation of Japan. Andrews, Edmund L. "Daimler Anticipates A Hard Year, Losing Perhaps $1.5 Billion," New York Times, February 27, 2001, pp. CI-C2. 27. Andrews, 2001. 28. Hyde, Justin. "Daimler faces tough choices in sharing parts," Reuters, February 27, 2001, pp. 1-3. 29. Ibid. 30. Andrews, 2001. 31. Hyde, 2001. 32. Andrews, 2001. 33. Ibid. 34. Robertson and Ulrich, 1998. Chapter 2 1. Ulrich, Karl. "The role of product architecture in the manufacturing firm," Research Policy, 24, 1995, pp. 419-440. 2. Ibid. 3. This example appears in Charles H. Fine's book Clockspeed: Winning Industry Control in the Age of Temporary Advantage. (Reading, Massachusetts: Perseus Books, 1998.) 4. Fine, 1998. 5. This example also appears in Charles H. Fine's book Clockspeed: Winning Industry Control in the Age of Temporary Advantage. (Reading, Massachusetts: Perseus Books, 1998.) 6. Fine, 1998. 7. Robertson, David and Ulrich, Karl. "Planning for Product Platforms," Sloan Management Review, Summer 1998, pp. 19-31. 8. Ibid. 9. Wheelwright, Steven C. and Clark, Kim B. "Creating Project Plans to Focus Product Development," HarvardBusiness Review, March-April 1992, pp. 70-82. 10. Robertson and Ulrich, 1998. 11. Meyer, Marc H. and Lehnerd, Alvin P. The Power of Product Platforms. New York, New York: The Free Press, 1997. 108 12. Pedersen, Per Elgaard. "Organisational Impacts of Platform Based Product Development," InternationalConference on EngineeringDesign, Munich, Germany. August 24-26, 1999, pp. 1507-1512. 13. Pulkkinen, Antti, Lehtonen, Timo, and Riitahuhta, Asko. "Design for Configuration Methodology for Product Family Development," InternationalConference on Engineering Design, Munich, Germany. August 24-26, 1999, pp. 1495-1500. 14. Martin, Mark V. and Ishii, Kosuke. "Design for Variety: Development of Complexity Indices and Design Charts," ASME Design EngineeringTechnical Conferences, Sacramento, California. September 14-17, 1997, DETC97/DFM-4359. 15. Kota, Sridhar and Sethuraman, Kannan. "Managing Variety in Product Families through Design for Commonality," ASME Design EngineeringTechnical Conferences, Atlanta, Georgia. September 13-16, 1998, DETC98/DFM-565 1. 16. Stone, R., Wood, K. and Crawford, R. "A Heuristic Method for Identifying Modules for Product Architectures," Design Studies, Vol. 21, No. 1, January 2000, pp. 5-31. 17. Siddique, Zahed and Rosen, David W. "Product Platform Design: A Graph Grammar Approach," ASME Design EngineeringTechnical Conferences, Las Vegas, Nevada. September 12-16, 1999, DETC98/DTM-8762. 18. Rechtin, Eberhardt and Maier, Mark W. The Art of Systems Architecting. Boca Raton, Florida: CRC Press, 1997. 19. Ericsson, Anna and Erixon, Gunnar. Controlling Design Variants: Modular Product Platforms. Dearborn, Michigan: Society of Manufacturing Engineers, 1999. 20. Moore, William L., Louviere, Jordan J., and Verma, Rohit. "Using Conjoint Analysis to Help Design Product Platforms," Journalof ProductionInnovation Management, 16, 1999, pp. 27-37. 21. Gershenson, J. K., Prasad, G. J., and Allamneni, S. "Modular Product Design: A Life-Cycle View," Transactions of the Societyfor Design and Process Science, Vol. 3, No. 4, December 1999, pp. 13-26. 22. Coulter, Stewart L., McIntosh, Mark W., Bras, Bert, and Rosen, David W. "Identification of Limiting Factors for Improving Design Modularity," ASME Design Engineering Technical Conferences, Atlanta, Georgia. September 13-16, 1998, DETC98/DFM-5659. Chapter 3 1. Stone, Robert B. and Wood, Kristin L. "Development of a Functional Basis for Design," ASME Design Engineering Technical Conferences, Las Vegas, Nevada. September 12-15, 1999, DETC99/DTM-8765. 2. Lawrence D. Miles is generally credited with developing the technique known as value engineering. Much of his work on value engineering can be found in his first book, Techniques of Value Analysis and Engineering (New York, New York: McGraw-Hill, 1961). Akiyama, Kaneo. Function Analysis: Systematic Improvement of Quality and Performance. Cambridge, Massachusetts: Productivity Press, Inc., 1991. 3. Heller, Edward D. Value Management: Value Engineering and Cost Reduction. Reading, Massachusetts: Addison-Wesley Publishing Company, 1971. 109 4. Akiyama, 1989. 5. Otto, Kevin N. and Wood, Kristin L. Product Design. Englewood Cliffs, New Jersey: Prentice Hall, 2001. 6. Akiyama, 1989. 7. Ibid. 8. Bischoff, W. and Hansen, F. Rationelles Konstruieren. KonstruktionsBUcher Bd. 5, Berlin: VEB-Verlag Technik, 1953. Rodenacker, W. G. Methodisches Konstruieren. KonstrktionsBticher Bd. 27, Berlin: Springer, 1970. Roth, K. Konstruieren mit Konstruktionskatalogen. Berlin: Springer, 1982. 9. Pahl, G. and Beitz, W. Engineering Design: A Systematic Approach. London: Springer- Verlag, 1996. 10. Otto and Wood, 2001. 11. Ibid. 12. Otto, Kevin N. and Wood, Kristin L. "Product Evolution: A Reverse Engineering and Redesign Methodology," Research in EngineeringDesign, 10(4), 1998, pp. 226-243. 13. Otto and Wood, 2001. 14. Stone, R., Wood, K. and Crawford, R. "A Heuristic Method for Identifying Modules for Product Architectures," Design Studies, Vol. 21, No. 1, January 2000, pp. 5-31. Chapter 4 1. Ishii, Kosuke. Course Reader ME217A: Design for Manufacturability: Product Definition. Stanford, California: Stanford University Bookstore, 2001. 2. Eubanks, C., Kmenta, S. and Ishii, K. "Advanced Failure Modes and Effects Analysis Using Behavior Modeling," ASME Design EngineeringTechnical Conferences, Sacramento, California. September 14-17, 1997, DETC97/DTM-3872. Reliability Analysis Center. Course Material: Mechanical Reliability Training Program. Rome, New York: IIT Research Institute, 2000. Stamatis, D. Failure Modes and Effects Analysis. Milwaukee, Wisconsin: ASQC Quality Press, 1996. 3. Gershenson, John and Ishii, Kosuke. "Life-Cycle Serviceability Design," ASME Design EngineeringTechnical Conferences, Miami, Florida. September 22-25, 1991, DE-Vol. 31, DTM, pp. 127-134. Gershenson, John and Ishii, Kosuke. "Design for Serviceability," in Concurrent Engineering: Theory and Practice, Kusiak, A. (ed.) New York, New York: John Wiley & Sons, Inc., 1992. 4. Ishii, Kosuke. "Life-Cycle Engineering Design," JournalofMechanical Design. Vol. 117, June 1995, pp. 42-47. 5. Newcomb, Patrick J., Bras, Bert, and Rosen, David W. "Implications of Modularity on Product Design for the Life Cycle," ASME Design EngineeringTechnical Conferences, Irvine, California. August 18-22, 1996, DETC96/DTM-1516. 110 6. Reliability Analysis Center, 2000. 7. Gnedenko, Boris and Ushakov, Igor. Probabilistic Reliability Engineering. New York, New York: John Wiley & Sons, Inc., 1995. 8. Siddall, James N. Probabilistic Engineering Design. New York, New York: Marcel Dekker, Inc., 1983. 9. Misra, K. B. Reliability Analysis and Prediction. Amsterdam, The Netherlands: Elsevier, 1992. 10. Ibid. 11. Lewis, E. E. Introduction to Reliability Engineering. New York, New York: John Wiley & Sons, Inc., 1996. 12. Ibid. 13. Moss, Marvin A. Designing for Minimal Maintenance Expense. New York, New York: Marcel Dekker, Inc., 1985. 14. Dudewicz, Edward J. "Basic Statistical Methods," in Juran's Quality Control Handbook, Juran, J. M. (ed.) New York, New York: McGraw-Hill, Inc., 1988. 15. Barkan, Phillip and Hinckley, C. Martin. "The Benefits and Limitations of Structured Design Methodologies," ManufacturingReview. Vol. 6, No. 3, September 1993, pp. 211-220. 16. Xerox Corporation. Xerox Docucolor 2000 Series Brochure. 2000. Chapter 5 1. Sanderson, Susan W. and Uzumeri, Mustafa. Managing Product Families. Chicago, Illinois: Irwin, 1997. 2. Otto, Kevin N. and Wood, Kristin L. Product Design. Englewood Cliffs, New Jersey: Prentice Hall, 2001. 3. Tushman, Michael and Anderson, Phillip. "Technological Discontinuities and Organizational Environments," Administrative Science Quarterly,31, 1986, pp. 439-465. 4. Christensen, Clayton M. The Innovator's Dilemma. Cambridge, Massachusetts: Harvard Business School Press, 1997. 5. Christensen, Clayton M. "The Opportunity and Threat of Disruptive Technologies," Presentation at Xerox, Inc., Rochester, New York. June 12, 2000. 6. Ulrich, Karl T. and Eppinger, Steven, D. Product Design and Development. Boston, Massachusetts: Irwin/McGraw-Hill, 2000. 7. Ibid. 8. Ishii, Kosuke, Juengel, Cheryl, and Eubanks, C. Fritz. "Design for Product Variety: Key to Product Line Structuring," ASME Design EngineeringTechnical Conferences, Boston, Massachusetts. 1995, DE-Vol. 83, DTM, pp. 499-506. Martin, Mark V. and Ishii, Kosuke. "Design for Variety: A Methodology for Understanding the Costs of Product Proliferation," ASME Design EngineeringTechnical Conferences, Irvine, California. August 18-22, 1996, DETC96/DTM- 1610. 111 Martin, Mark V. and Ishii, Kosuke. "Design for Variety: Development of Complexity Indices and Design Chart," ASME Design Engineering Technical Conferences, Sacramento, California. September 14-17, 1997, DETC97/DTM-4359. 9. Ericsson, Anna and Erixon, Gunnar. Controlling Design Variants: Modular Product Platforms. Dearborn, Michigan: Society of Manufacturing Engineers, 1999. 10. A conjoint analysis can be used to help estimate the relative importance that customers place on technology, cost, and other attributes for any given product. The conjoint analysis method is well described in the following article: Green, Paul E. and Wind, Yoram. "New Way to Measure Consumers' Judgements," HarvardBusiness Review, July-August 1975, pp. 107- 117. 11. An example of this is the Global Positioning System (GPS) developed by the US Government. When GPS finally reached its full operational capacity in 1995, the government, in the interest of national security, intentionally degraded the GPS signal so as to limit the accuracy for civilian users. In this case, while the technology existed, it was not provided to the public with its full functionality. On May 1, 2000, the US Government stopped degrading the GPS signal. 12. Much of the information used in this example was made up for the purpose of demonstrating the methodology. The information used is not based on actual data and does not reflect actual information from Xerox. Chapter 6 1. Fine, Charles H. Clockspeed: Winning Industry Control in the Age of Temporary Advantage. Reading, Massachusetts: Perseus Books, 1998. 2. Ibid. 3. Fine, Charles H. and Whitney, Daniel W. "Is the Make-Buy Decision Process a Core Competence?" Working Paper, Massachusetts Institute of Technology, Cambridge, Massachusetts, February 1996. http://web.mit.edu/ctpid/www/Whitney/morepapers/makeab.pdf 4. This case is presented in greater detail in Charles H. Fine's book, Clockspeed: Winning Industry Control in the Age of Temporary Advantage. (Reading, Massachusetts: Perseus Books, 1998.) 5. Fine, 1998. 6. Davis, Tom and Sasser, Marguerita. "Postponing Product Differentiation," Mechanical Engineering,November 1995, pp. 105-107. Lee, H. L. and Tang, C. S. "Modeling the Costs and Benefits of Delayed Product Differentiation," Management Science, Vol. 43, No. 1, 1997, pp. 40-53. Ulrich, Karl T. and Eppinger, Steven, D. Product Design and Development. Boston, Massachusetts: Irwin/McGraw-Hill, 2000. 7. Davis and Sasser, 1995. 8. Ericsson, Anna and Erixon, Gunnar. Controlling Design Variants: Modular Product Platforms. Dearborn, Michigan: Society of Manufacturing Engineers, 1999. 112 9. Ishii, Kosuke. "Modularity: A Key Concept in Product Life-Cycle Engineering," Working Paper, Stanford University, Stanford, California, 1998. http://mml.stanford.edu/Research/Papers/1998/1998.LEbook.ishii/1998.LEbook.ishii.pdf 10. http://www.mabuchi-motor.co.jp/english/index.html http://www.johnsonmotor.com http://www.winston.com.hk 11. http://SDP-SI.com http://www.wmberg.com http://www.pic-design.com/index.htm 12. Newark Electronics, Catalog 119, 2001. 13. Ibid. Chapter 7 1. Clausing, Don. Total Quality Development. New York, New York: ASME Press, 1994. Gonzalez-Zugasti, Javier P. and Otto, Kevin N. "Platform-Based Spacecraft Design: A foundation and Implementation Procedure," IEEE Aerospace Conference, Big Sky, Montana, 2000. Vol. 1, pp. 455-463. Hollins, Bill and Pugh, Stuart. Successful Product Design: What to do and When. London, England: Butterworth & Co. Ltd., 1990. Otto, Kevin N. and Wood, Kristin L. Product Design. Englewood Cliffs, New Jersey: Prentice Hall, 2001. 2. Antonsson, E. K. and Otto, K. N. "Imprecision in Engineering Design," Journalof MechanicalDesign, Vol. 117, June 1995, pp. 2 5 -3 2 . Thurston, D. "A Formal Method for Subjective Design Evaluation with Multiple Attributes," Research in EngineeringDesign, 3(2), 1990, pp. 105-122. 113