Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
THE STRATEGIC APPROACH TO - PRODUCT DESIGN DANIEL E. WHITNEY, JAMES L. NEVINS, THOMAS L. DE FAZIO, RICHARD E. GUSTAVSON, RICHARD W. METZINGER, JONATHAN M. ROURKE, AND DONALD S. SELTZER ABSTRACT The strategic approach to product design (SAPD) is a multistep pro- cess that seeks to implement integrated product and process design. Because of the inherently integrative nature of the assembly process, that is Me focus of SAPD. This paper outlines the steps of SAPD, compares it to conventional product design methods, and suggests research that is needed to provide analytical and computer support to what is at present a team approach dependent on experts. INTRODUCTION There is growing concern that U.S. man- ufacturing is no longer competitive with that of other countries in the global market- place (National Research Council, 1986~. Many causes for this concern have been presented (Olmer, 1985; Porter, 1986; Wheelwright and Hayes, 198S), and many remedies have been proposed. The impor- tant point is that, "despite the enthusiastic claims of technology developers and ven- dors, technology alone will not improve competitiveness" (National Research Coun- cil, 1986, ix). What is needed is an inte- grated approach to manufacturing systems supported by new technologies. This paper discusses a strategic approach being used by some U.S. and many Japanese companies to raise their productivity by both integrat- ing their design and manufacturing func- 200 tions and modifying their manufacturing institutions. A new generation of advanced manufac- turing technology, most notably flexible manufacturing systems (FMSs), is having an important impact on unit operations. This is particularly true of product design that is carried out through CAD systems that are part of an EMS. Although such systems allow many operations to be car- ried out automatically, the impact of these new approaches on the functional way in which most companies operate has been minimal. These conditions can be contrasted with what is occurring when companies consider the next generation of assembly, namely, automated flexible assembly. There is a growing awareness that automation for as- sembly cannot be treated in an isolated manner. Assembly, with its close coupling
THE STRATEGIC APPROACH TO PRODUCT DESIGN to design, vendor control, quality control, etc., requires a new, more highly inte- grated approach to manufacturing. Assem- bly is inherently integrative. Parts that were separately designed, made, handled, and inspected must be joined together, handled together, and tested together. The assembly process focuses attention on pairs and groups of parts. Decisions that affect assem- bly also affect nearly every other aspect of production and use of a product. Assembly, therefore, is a natural forum for launching an integrated attack on all the phases of a product, from conception and fabrication to quality and life cycle. The recent interest in design for assembly is an outgrowth of the realization that assembly is an impor- tant phase in the life cycle of a product. Other phases, however, are also important. They may last longer and cost more. Thus, design for assembly cannot be done in iso- lation. An approach that integrates product de- sign and all aspects of the manufacturing process can only be accomplished through a strategic approach to product design (SAPD). Such an approach allows the en- tire system to be rationalized. Further- more, it helps identify the need for CAE data bases and computerized design tools that are needed to support this tightly inte- grated activity. SAPD provides the opportunity to deal with the many trade-offs that must be made. It also allows it to be done at the best time, namely, when the product is being designed. Although design for assembly usually deals with single parts, SAPD deals with groups of parts, subassemblies, and the total product. SAPD is independent of any particular assembly technology. It deals with many nonassembly issues, and it provides ratio- nally designed products. This rationality benefits assembly by any technique, in- cluding manual; it benefits other phases of production, such as inspection; and it 20] can materially affect the life cycle of the product. Although many companies recognize the advantages of an integrated approach to product design, Japanese firms have been the most proficient in using it, some for 20 or 30 years. A number of advanced U.S. companies have recently embraced this ap- proach. The remainder of this paper provides two views of SAPD. The intuitive approach, as it is usually practiced in both the United States and Japan, is described first. Second, the intellectual ingredients of this approach are extracted and listed so that the research problems can be seen more clearly. Finally, some of the educational, curricular, and technology transfer implications are dis- cussed. Before the process is described in detail, it is important to give some of the context. DRIVERS OF CHANGE IN MANUFACTURING i] Part fabrication is essentially a series of Dependent steps with minimum intercon- nections between operations being per- formed on individual piece parts. These op- erations are intended to enforce a particular geometric configuration on formable mate- rials. The ideal geometry exists in the de- sign. Because of statistical variations in both the process and the materials, the fabrica- tion process creates a part whose properties only approximate those of the ideal. These perturbations must be controlled so that the resulting part has properties that fall within design tolerances. Models for fabrication processes are gen- erally good. Thus, the degree to which un- certainty perturbs the ideal geometry can be predicted fairly well certainly well enough to allow automatic fabrication to proceed with high confidence. The level of understanding of assembly and inspection, however, is much less mature.
202 The technology of part fabrication has advanced considerably in the past 30 years. In the early days, single numerically con- trolled machines took their instructions from hand-delivered paper tapes. Today, systems of 10 or 20 such machines work together to make groups of parts that were designed on computers. The tapes have been replaced by direct data links. The de- signs are now supported by three-dimen- sional geometric modeling programs and by computers that hold information on such things as materials, stress analysis, and ma- chining methods. Engineers and research- ers currently face a problem of logistics and scheduling that is, finding the right mix of parts to keep the machines busy. Few tools are currently available to aid in solv- ing these problems. The production capacity of such auto- mated systems is impressive, and so is the speed with which parts can be designed and made. Based on the performance of these design and fabrication systems, most man- ufacturers are convinced of the benefits of close integration between design and fabri- cation. However, too few manufacturers recognize that design for fabrication is not the same as design for overall producibility. Design for fabrication, sometimes mistak- enly called design for manufacturability, considers parts as isolated entities rather than in groups that must function together. SAPD seeks to correct that shortcoming. Beyond this, several deeper problems have become visible. The logistical prob- lems mentioned earlier have sensitized manufacturers to two needs: to integrate scheduling and system operation into their strategies, and to design parts and families of parts to make better use of these systems. Existing systems are not adequately flexi- ble. They are suited to only one kind of manufacture, that of serial metal removal or metal bending on a limited variety and on a small number of carefully chosen pieces. Other kinds of manufacture, such as WHI TNEY E T AL. casting, composite material lay-up, and powder metallurgy, and other steps besides fabrication, such as assembly, have not yet been brought under the umbrella of computer-integrated systems. Extensions like these require much more careful prod- uct design and process organization. The response to these challenges has been a surge of interest in DFA. Ten or twelve years ago the "correct" attitude of automa- tion engineers was to take the product as it was designed and do their best to assemble it, either manually or by machine. Manu- facturers tended to emphasize low part cost and fast assembly. The result was that all the problems introduced by fabrication or logistics, including parts out of tolerance, or late, or damaged, had to be solved by the ingenuity of the assemblers. This was the social structure of manufacturing a sort of hidden agenda. This hidden agenda made the introduc- tion of robots and other advanced assembly technology either difficult or impossible. Manufacturers were reluctant to pay for the cost of extra equipment to support the ro- bots, such as part feeders, palletized parts, and control computers. They were also dis- appointed that robots were not as resource- ful as people in handling problems such as out-of-tolerance parts. Few companies wanted to spend the extra money for higher quality parts just so robots could put them together. Instead they demanded better and more intelligent, adaptive robots that could solve these problems. In the past few years these attitudes have begun to change. Manufacturers are real- izing that, although better quality parts make possible assembly by robots, they also create a better quality product. They are realizing that using the assembly system as a filter to detect bad parts is a bad way to run a factory. The hidden agenda is grad- ually being stripped away. With this is coming a deeper understanding of the role of technology in products and processes. In
THE STRATEGIC APPROACH TO PRODUCT DESIGN addition, we see the convergence of forces that will demand new approaches. These forces are · The complexity of new products and the disappearance of the learning curve; · The complexity of modern worldwide production and the changing nature of competition; and · The disappearance of bly as an option. manual assem- Complexitr of Products and the Disappearance of the Learning Curve Modern products can contain thousands of parts and many technologies. A new au- tomobile can take more than 5 years from initial specifications to production. A new surface combat ship probably the most complex item built today can take up to 10 years to cover the same process. Modern products are characterized by combinations of energy and information storage systems. They may be made of materials that are not merely transformed from mined ores or feedstocks but are created with properties that especially serve the needs of the prod- uct. Some new products are tailored to spe- cific market niches, thus demanding small production volumes. Thus, there is a need for rapid advances in a product line, fast updating of designs, and quick changes in production schedules. The learning curve that allowed design or production problems to be worked out over a long period of time has been compressed or eliminated. In- stead, learning must be spread over a series of products. To be valuable, this learning must focus on generic issues rather than product-specific issues so that the lessons can be passed on to the next product. Two re- sponses to the disappearance of the learning curve can be identified. First, manufacturers have responded to the time compression by doing things faster. Products and systems are being de- 203 signed more rapidly, in part through better planning and more effective computer tools. Better planning means organizing the de- sign process so that more factors are taken into account early, reducing the chance of damaging surprises later. It also involves identifying a good sequence in which to make design decisions, thus retaining some room to maneuver in the later design stages and also permitting decisions to be made in a way that minimizes the need to iterate (Akagi et al., 1984~. Better and more com- prehensive computer tools, computer-aided- design, computer-aided engineering, and computer-aided manufacturing allow cal- culations to be made more rapidly and to cover more cases. Second, manufacturers must think more deeply about what they do and the most effective use of the lessons learned from pre- vious design activities. Although the learn- ing may still take longer than the time available to design a new product or sys- tem, learning can occur on a continuous basis. The basis for the systemization is the recognition of generic or repeating elements in the product, the processes, or the design steps themselves. An analogous approach, called group technology (Opitz, 1967), may be found within the narrowly defined topic of ma- chined parts. Although the application of the approach differs from case to case, the spirit of it is to recognize major similarities or differences between nonidentical items so that they can be grouped. Within a group, the similarities are used to advan- tage by means of sharing, for example, a machine or a measuring method. When fa- cilities can be shared by a large number of items, time and cost are usually saved. Had the similarities not been recognized and the groups not formed, the facilities would have dealt with the items piecemeal, and no quantity savings could have been made. To apply this idea in design, one needs a systematic, step-by-step approach that asks
204 the same kinds of questions and requires repeated applications of the same kinds of analyses. As these design steps are repeated, one should become better at recognizing the similarities, even though each product or system is outwardly different. As the steps become clearer, it should be possible to de- velop computer tools for carrying them out. A major aim of this paper is to highlight these steps and identify both the existing tools and the gaps where more knowledge is needed. The formulation of these steps and their organization into a coherent ap- proach constitute the intellectual challenge of manufacturing. Complexity of Processes and Me Changing Nature of Competition Production processes are complex. Since many products are changing frequently, timing is critical. The result is that processes require more care and attention, and more data are needed to determine how a man- ufacturing system is performing. New pro- duction technology requires new skills and attitudes from workers and managers. Product and process complexity arise from the appearance of new kinds of prod- ucts, many of which contain a true mix or integration of mechanical and electronic functions, thus requiring more broadly ed- ucated product and process designers. An example of such a product is computer disk drives, whose success depends on careful design, precise tolerances, extreme cleanli- ness, fine timing and balance, and the skill and attention of dedicated people. Successful competition for markets for such products demands a new response. In older industries or in those that have reached maturity, the basis of competition is usually production efficiency. The man- agers in such industries focus on asset man- agement and make most of their decisions based on incremental economic criteria. In WHI TNEY E T AL. newer industries, or in older ones faced with new competitors, the bases of competition are more likely to be product innovation, advanced technology, and quality. A1- though mature products probably differ lit- tle in technology and are distinguished by price, newer products may command a price premium based on quality or novelty. Manufacturing decisions in such industries are therefore less likely to be dominated by incremental economics and more by the ability of the product and factory to sup- port the competitive strategy behind the product, such as the ability to evolve rap- idly or be responsive to a changing market while maintaining a high product quality. Disappearance of the Manual Assembly Option Since manual metal removal was never an option, machinery for this purpose de- veloped early in the industrial revolution. Thus, metal removal processes were among the first to be well understood. While man- ual assembly has been a common past prac- tice, it is rapidly disappearing as an option in high-technology products. People have too much difficulty providing the required quality, uniformity, care, documentation, and cleanliness that are required of many of today's products. This is not to say that remarkable human performance is impos- sible. In Japan, for example, it is usual for a worker to make only one assembly error (wrong, missing, or broken party in 25,000 to 100,000 operations. Since this is consid- ered not good enough, the Japanese use these error numbers to justify further auto- mation. Such is their conviction that mod- ern products demand even higher quality. . . Direct substitution of robots for people will not solve the problem, however. The successful use of robots requires a carefully designed environment consisting of a prop- erly designed product, well-trained opera-
THE STRATEGIC APPROACH TO PRODUCT DESIGN tors, and well-scheduled operations. Design THE STRATEGIC APPROACH TO for assembly Is also not enough by Itself. PRODUCT DESIGN Since many parts or subassemblies will be purchased, the same kinds of problems must be solved by outside vendors. The decision to replace a person with a piece of machinery, such as a robot, always involves some form of economic analysis. The commonly used analytic techniques as- sume that the substitution results in a sys- tem that is equivalent and interchangeable with the current system in every way except cost. Other factors need to be included, however. Since each proposed replacement is different in its ability to deliver quality, it is important that this be reflected in the analysis. Failure rates, repair costs, and testing strategies, for example, must be con- sidered. These in turn are affected by prod- uct design, as discussed later. In high-tech- nology products, the cost of materials is frequently more important than time or la- bor, so the ability of an assembly-test sys- tem to deliver a good yield is crucial to maintaining production volume and profit. Thus, economically justifiable manufactur- ing systems can contain both machines and people. The techniques of economic analy- sis must also be improved to the point that a proper assessment can be given of the true value of fully integrated assembly systems and new product designs. Economic mea- sures, such as return on investment and ma- chine replacement, must be supplanted by more sophisticated criteria. All this means that we have to under- stand assembly as thoroughly as we now understand metal cutting. Indeed, all the processes in manufacturingmaterial han- dling, stocking, transport, inspection, judg- ment of suitability, and granting of "excep- tions" that are now routinely handled by people in an intuitive, judgmental, and of- ten undocumented way, will have to be so on. brought to a higher level of understanding, even if they are not to be executed by machines. 205 The way a product traverses its life cycle, including fabrication, purchase, assembly, inspection, use, repair, modernization, and disposal, is established when the product is designed. The effectiveness, efficiency, and cost of these various stages are all affected by decisions made during the design, which includes both product and process design. Since the product design must recognize strategic issues related to both the manufac- turer and the user of the product, the re- quired response has been called the strate- gic approach to product design. The remainder of this paper focuses on SAPD as a process and as a discipline. The Educational Problem Engineering schools teach a fairly straightforward version of how something is designed. This version of the design pro- cess is shown schematically in Figure 1. En- gineers are given a technically oriented view that begins with the need for the product; proceeds to the preparation of product specifications, the making of trial designs, prototypes, and final designs; and con- cludes with a manufacturing process plan. There is a good deal of feedback as prob- lems are uncovered and resolved. But in the main the process is self-contained from need to final design, with little outside inter- ference. This method suffers from too much line- arity of the process, it is often too techni- cal, and it is too compartmentalized, en- couraging design to be the domain of the designer, manufacturing the domain of the manufacturing engineer, purchasing the domain of the purchasing manager, and A greatly improved method, one used by companies that are more successful in creating competitive products, shares at-
206 Production System Design | Production ~.,M_ I FIGURE 1 The conventional product design- production system design process. tributes with the procedure shown in Fig- ure 2. This method emphasizes the degree to which decisions made by the different parties affect other activities and alter the product's character (National Research Council, 1986, 102-112~. An example will illustrate the type of problems that may arise when one unit is unaware of the needs imposed by other units on the design of a product. Consider a particular military product that depends critically on an infrared detector. The pur- chasing department switched to a lower- price vendor without determining the re- producibility of the detectors that would be provided. Although subtle differences be- tween detectors can significantly affect per- formance, these cannot be found until the product is partially assembled with optics, power supplies, and so on. To increase rug- gedness and reduce cost, the unit is glued together, making disassembly to replace the detectors very expensive. Naturally, the product could be redesigned with threaded WHITNEYETAL. l~larket Needs I Product L P - ormance Spece Product Deal9n joints to facilitate detector replacement as well as field repair. But the product in question is a single-use weapon. It must work the first time, its operating lifetime is only a few seconds, and its shelf life must be several years. Repair is simply not "in character" for this kind of an item. Thus, a decision to reduce costs by substituting one component for another can have conse- quences that can be appreciated only if the Production Production Inve~nent ~ life history of the product is completely un- Sy~tem ~ system ~ Decl~lon ~ Technob9y Spoca hleti~od. ~ derstood. The point of this example is that a seem- ingly minor decision, made to optimize a corner of a company's operations, can have a pervasive effect on how a product is made or how it performs in the field. These deci- sions can completely defeat the designer's intentions. Management, engineering, pur- chasing, personnel, and manufacturing can each contribute to making or defeating a product. Market Needs Production Syatern Tedilobgy | Product | | Performance | 1 SPEW 1 1 ~~:~ 1 ,.. ~ 1 Investment Decision MetilOdS Product Design Production Hem I Prodllction L Cost Model FUGUE 2 The emerging concurrency method of designing products and production systems.
THE STRATEGIC APPROACH TO PRODUCT DESIGN Levels of Product Design Strategies Product design can be divided into levels of activities that include functional design, manufacturing, and life-cycle considera- tions. Product designers traditionally do functional design. They choose materials, dimensions, and tolerances in such a way that the item will accomplish its intended purpose. In traditional organizations, the functional design is given to the manufac- turing engineers to determine the processes for fabricating each part, including choice of machines, methods for maintaining tol- erances, and make-or-buy decisions. The latter decisions essentially export to vendors some of the manufacturing engineers' prob- lems. For this export to be successful, some- one (e.g., the engineers or the purchasing agents) must carefully monitor the vendors. Irrational products and production systems can result if the monitors do not sufficiently understand the product and its require- ments. In addition, the manufacturing engi- neers must design the assembly system or method for the product. Traditionally, this is done by straightforward economic anal- yses, in which a choice is made between manual and"automatic" assembly. The latter is chosen usually in cases where the product is small, has less than 12 parts or so, and is made in quantities of about a million per year for several years. In recent years, the need for improved competitiveness and productivity has led many companies to modify this process to recognize the entire life cycle of the prod- uct. When viewed in this manner, the pro- cess comprises life cycle issues such as prod- uct use, repair, and upgrading, manu- facturing issues such as the assembly sys- tem, assembly operations, tolerances, and vendor control; fabrication issues such as make or buy and method of fabrication; and functional design issues. One conse- quence of this new view is that manufactur- 207 ing engineers are involved earlier in the design. Unfortunately, neither the more tradi- tional nor the more recent sequence ade- quately describes practice. Converting a concept into a product is a complex proce- dure of many steps. As the design evolves, choices must be made concerning such things as materials, fasteners, coatings, ad- hesives, and electronic adjustments. Not only are these choices interdependent, it is likely that some of them would have been different if slightly different criteria for choice had been used by the designers. Fur- thermore, it becomes increasingly difficult as design proceeds to introduce new view- points and criteria without seriously delay- ing the design. Thus, if manufacturing en- gineers participate in the design debate from the start, their criteria can be properly considered. Similarly, if repair engineers, purchasing agents, and other knowledge- able people are represented, a better, more integrated design will result. In each case, the design will represent an interconnected web of decisions, and the participation of more parties will ensure that the web is better balanced. To bring some structure to a detailed de- scription of strategic design, the following breakdown of topics is used: · The character of the product is deter- mined; this identifies the attributes of the product. · A study of the design for producibility and usability is carried out. · A product function analysis is made to determine if the product's producibility and usability can be improved without impair- ing desirable functions. · An assembly process is designed that includes a suitable assembly sequence, the identification of subassemblies, the integra- tion of a quality control strategy with as- sembly, and the design of each part so that its functional tolerances and tooling toler-
208 ances (gripping and jigging surfaces) are compatible with the assembly method and sequence. · A factory system is designed that fully involves the production workers in the pro- duction strategy, operates on minimum in- ventory, and is integrated with the methods and capabilities of the vendors. Part fabrication, although essential, is not discussed in detail in this paper because it is a much more mature process. Given the present state of design meth- odology and techniques, the best way to pursue these activities is to form teams of product designers and manufacturing engi- neers, with active participation by repre- sentatives from marketing, finance, pur- chasing, and personnel. Some companies call this process top-down analysis; others call it concurrency. To be successful, a con- currency team should be formed early and maintained in position until the product is test-marketed. Figure 3 sketches the activities of the con- Character of the Product Preliminary Design Plus Economic Goals Product Funt Ton Analysis rat ~ Examine ~ em/ Part and \ \ Each Part ~ Fixture / ~ ~ I ~ ~ ~ ~ Improve ~ ~ / {Group ~ Robustness Alternate ~ =/ ~~~ Assembly | ,? Sequences j; / ~ Assembly Slmpilty \< ~' System Product ~ / \ r ~.,l^~-t \ Deslgin / 1 ~:V~, Estabilah Test and ac | Recomrn end l | Product Design l FIGURE 3 Outline of the strategic approach to product design, with emphasis on assembly is- sues. {Ope~aUonal~ {Economic,\ ~ ~ t~ WHITNEY ET AL. currency team (omitting fabrication issues) and shows how the activities interact. The next few sections of the paper expand on some of these activities. Character of the Product A product's character is the combination of the basic features of how the product will be made, sold, and used. It must be deter- mined early and recognized by everyone involved in the process. There is an endless list of possible features contributing to the character of a product. The following is an example of two product characters, to- gether with their consequences for product design and production: CHARACTER OF PRODUCT 1. Complex item, no model mix, used by untrained people, must have 100 percent reliability, used only once and thrown away. CONSEQUENCE Make high-quality parts, glue them together, do not try to fix after man- ufacture. 2. Complex item, contains a mixture of models and options, used by untrained peo- ple, lasts for years, and is serviced in the field. CONSEQUENCE Make high-quality parts, screw them together, and provide replacement parts and field repair service. Design for Producibility and Usability Once the character of the product has been defined, at least provisionally, true product design can begin. Design for pro- ducibility and usability is a top-down pro- cess. It is guided by the product, and it helps formulate the manufacturing strat- egy. This contrasts with many so-called ex- amples of design for assembly which are in fact just good (sometimes very good) re- engineering of the product itself without regard to an overall strategy. Innovative
THE STRATEGIC APPROACH TO PRODUCT DESIGN engineers can always come up with "im- provements." Without a guiding strategy, there is no way to tell which improvements really support the strategy and which merely look like isolated improvements. The main targets of the concurrency team are to · Convert the product concept into a manufacturable, saleable, usable product design. · Anticipate fabrication and assembly methods and problems. · Simplify the design, fabrication, use, and repair by, for example, reducing the number of parts or identifying and increas- ing the number of parts common to differ- ent models. · Improve the robustness of product and process by, for example, breaking product and process into self-contained modules, adjusting tolerances to eliminate chance failures, and identifying places where tests can be made. It is readily apparent that design for producibility is different from value engi- neering, an activity aimed chiefly at reduc- ing manufacturing cost by astute choices of materials or methods of making parts. Value engineering occurs after major prod- uct design is finished, and thus it is neither concurrent nor likely to be very thorough. The required thoroughness cannot be ac- complished except through a concurrent process. More importantly, design for pro- ducibility includes and sometimes subor- dinates reduction of manufacturing costs within the larger goal of optimizing the en- tire life cycle of the product. The reason is that, although employees assemble it, cus- tomers or repair personnel may disassemble it, and the actions of these others can be made easier, safer, or more congenial to the character of the product by decisions made simultaneously with design or assembly de- cisions. Design for producibility is also different from design for assembly. This activity, like 209 value engineering, usually begins after the product is designed. It considers the parts one by one, simplifies them, combines some to reduce part count, or adds features to make fabrication or assembly easier. This can be characterized as a bottom-up pro- cess. It is guided by the parts rather than by a holistic concept of how the product is to be made and used. Since the essence of a sophisticated design can depend on the careful choice of toler- ances, materials, or novel fabrication meth- ods that cannot be separated from the de- sign of the manufacturing process, the concurrency process must begin early. In- deed, in some cases, the process is the de- sign. An example of this process comes from Japanese shipbuilders. Their philosophy is that "design is a subset of production." Shipbuilding is an intensely complex and time-consuming process (Chirillo, 1982; Whitney et al., 1986~. The efficiency of shipbuilding is so heavily influenced by planning and organization that the Japa- nese have evolved a method that makes ac- tual design of the ship a part of the con- struction planning process. Since welded joints are just as strong as the surrounding metal, the shape of the pieces that are welded and the location of the joints can be chosen to facilitate producibility, providing a new freedom in design. The Japanese carefully choose subassembly and module shapes to exploit efficient group-technology methods for making them. Once the overall shape and characteris- tics of the ship have been determined, the size and shape of the pieces are determined by the order, method, or location in which they will be made. The size and shape of subassemblies into which those pieces will be built are first decided. Then the planners identify and give shape to the individual pieces of hull plate, pipe, deck, and so on, including the precise schedule of ordering raw material, joining of parts, and mea- suring to ensure that the assemblies will fit together the first time. Each of these sub-
210 assemblies is called a zone, and all manage- ment, scheduling, cost accounting, and su- pervision is done by tracking these zones through several predetermined stages of production. Zones at a particular stage are grouped into similar areas, where each area constitutes a type of work with similar needs for human skills, machinery, measuring equipment, and so on. The ship is designed so that the maximum possible number of zones constitute areas that are easy to make. Some examples, shown in Figure 4, indi- cate the levels of planning and production that the Japanese have introduced in par- allel with their new designs. The relation- ship of the builders to the steel mills allows them to order the precise shape of plates they need, on short notice, and with the necessary uniform quality that permits carefully developed low-distortion welding methods to be used. Many ships were built during the time that this procedure was being developed. Although ships differ in detail, the process not only emphasizes and takes advantage of the similarities but also encourages the de- signers to use these types of similarities and to identify rational ways of improving the producibility of later ships. The full potential of design for produci- bility cannot be realized until concurrency team members fully understand how the product is supposed to work and be used. They achieve this understanding through analysis of product function. Product Function Analysis Product function analysis is an activity in which designers and engineers seek ways of simplifying or rationalizing a product's de- sign by starting from what the product should do rather than how it performs that function nou) or how that function was per- formed in previous designs. Decisions re- garding fabrication or assembly method can affect users as well as factory personnel and field costs as well as factory costs. These WHI TNEY E T AL. decisions are related to design, not manu- facturing. Because they affect the character of the product, they are strategic in their impact. No one department should make these decisions alone, nor can the decisions be parceled out for decentralized action. These are decisions that the concurrency team must address. Assembly Processes The concurrency team must also address assembly processes. Activities with strategic implications include blies; · Division of the product into subassem- · Establishment of an assembly se- quence; · Selection of an assembly method for each step; and · Integration of a quality-control strat- egy. There is no set order in which to consider these activities, since the choices interact, and making them may trigger more design changes. It is convenient to discuss the first two of these activities together. The choice of assembly sequence and the identification of subassemblies focuses at- tention on so many aspects of product de- sign that they provide a natural starting point for integrative detailed design. As- sembly sequence studies require identifica- tion of potential jigging and gripping sur- faces, grip and assembly forces, clearances and tolerances, and other issues that must be accounted for in component design. A1- though these issues were not considered im- portant when manual assembly was used, they are very important to machine assem- bly. For example, tolerances on grip and jig surfaces must be adequate with respect to mating surfaces. Tolerance adequacy can be determined using the Part Mating The- ory that has emerged in recent years (CSDL Reports 1974-1980, R-800, R-850, R-921, R-996, R-1111; R-1276; 1979-1982, R-1407,
RLC,CK AS~FMRLY LFVFL Or FRAMINGSTAGE | ASSEMBLY STAGE l BLOCK ASSEMBLY LEVEL l ASSEMBLY STAGE it. , SEMI-BLOCK ASSEMBLY LEVEL ASSEMBLY STAGE \~> | GRAND BLOCK JOINING LEVEL l - 59.0 TONS INCLUDING OUTFIT / // ' ~ 31.8 TONS . . FIGURE 4 Example of division of ship hull into manufacturable subassemblies that may be grouped into like production classes. SOURCE: Chirillo (1982~. 211
212 R-1537; Whitney, 1982; Whitney et al., 1983~. In addition, sequence issues high- light assembly machine and tooling design problems, such as part approach directions, tolerance buildup due to prior assembly steps, access for grippers, stability of sub- assemblies, number of tools needed, and tool change requirements. Thus, the choice of sequence, normally considered late in the process design, really belongs in the early stages, since each can heavily affect the other. For these reasons, "determination of alternate assembly sequences" occupies the center of Figure 3. The benefit of this approach to assembly can be seen by considering a hypothetical product of six parts. It can be built in many ways, among them bottom up, top down, or from three subassemblies of two parts each. The choice among these options de- pends on many factors. There are construc- tion considerations, such as access to fasten- ers or lubrication points. There are assembly considerations, such as sequences whose success may be doubtful or whose failure might damage some parts. There are qual- ity control whose failure might damage some parts. There are quality control con- siderations, such as the ability to test the function of the subassembly before it is bur- ied beneath many other parts. There are process considerations, such as ability to hold the pieces accurately during critical operations. Finally, there are production strategy considerations, such as being able to make in advance some subassemblies that are common to many models, allowing fi- nal assembly to be completed quickly on the remaining parts. In traditional industrial engineering (Taylor, 1911), a major influence on choice of assembly sequence is line balance. Rele- vant to manual assembly, line balance is achieved by dividing up the assembly tasks so that each worker's total task time per cycle is as close to that of the other workers as possible. To achieve line balance, the WHITNEY ET AL. industrial engineer decides on sequences and groups the tasks and assigns the groups to the workers. It should be clear from the foregoing discussion that much broader is- sues can be brought to bear on sequence choices, ranging from testing options to market strategy. It should also be clear that a sequence that is good for human workers may be totally irrelevant for machines, whose strengths and weaknesses are totally different from those of people. For exam- ple, it is easy for a person to turn over a small item while passing it on to the next station. For a robot or machine to perform the turnover, it must be provided with an extra powered axis at a cost that may be considerable. This brings us to the choice of an assem- bly method for each step and integration of a quality control strategy. The greatest in- fluences on choice of method are the antic- ipated production volumes and the need for flexibility in model mix, part count, op- tions, method of treating units that fail tests, and so on. There is some literature on this topic, centering mainly on the technical as- pects and capabilities of different methods, but there are few detailed comparisons of accuracy or reliability. There is also a dearth of economic data. The result is that one cannot at this time make a convincing prediction of the cost and throughput of candidate assembly systems. In certain spe- cialized cases, good estimates can be made. These cases include manual systems or those consisting of specialized "assembly ma- chines." Even in these cases, there are few data and models, and most decision making is based on informed estimates by experi- enced individuals. This is a developing problem area, and much work remains to be done. A number of computer-based tools for designing and analyzing assembly systems have been developed by the Draper Labo- ratory (CSDL Reports 1978-1980, R-1284, R-1406; Graves and Lamar, 1983) and by
THE STRATEGIC APPROACH TO PRODUCT DESIGN others, notably Boothroyd et al. (1982~. Given adequate data, these tools permit the following issues to be addressed: · What is the best economic mix of ma- chines and people to assemble a given model mix of parts for a product, given each machine's or person's cost and time to do each operation, plus production rate and economic return targets? · How much can one afford to spend on an assembly system, given an anticipated Conclusion revenue stream? · How much extra time, machines, money, or product inventory are required . · · to meet a production rate it a certain mix of failures and repair steps can be antici- pated during production? · How can one make the trade-off be- tween the cost of higher-quality parts and the time to unjam an assembly machine when a low-quality part gets stuck? · What is the best way to distribute work among workstations in an assembly system? Integrating a quality-control strategy into product and process design involves many decisions, including purchasing options and personnel policies that are beyond the scope of this paper. We will touch on only two aspects related to topics already discussed: definition of subassemblies and modular assembly-line design. We have seen that a way to define subassemblies is to define assembly stages in which an object with a definable function has been built. Since that function is related to the product's specifi- cations, we should be able to define a test for that subassembly so that we know it will do its job when mated to the rest of the parts. To incorporate a quality control strategy into manufacturing, it is necessary to de- cide which of many possible assembly stages to choose as test points. Relevant factors include how costly and how definitive the test is, whether hidden flaws could become undetectable if the test were delayed to later 213 stages, and how much it costs to repair or discard bad subassemblies. It is not uncom- mon for a tear-down at final assembly to cost as much as half the total manufactur- ing cost. A rational approach involves ex- amining the assembly sequence to deter- mine where each fault becomes critical and each test opportunity occurs. This study may result in a new assembly sequence. Company management must establish a product strategy and encourage the design team to search for the product and manu- facturing system design that best suits the product's character and meets the needs of the marketing and manufacturing ap- proach. It should be clear that a great deal of work is required to think through the options and arrive at a good final design. Technology alone cannot create a produc- tive manufacturing organization. Fortu- nately, the emergence of new system design tools, methods, and part mating models and analyses has created a new knowledge base to support this work. Furthermore, institu- tional changes will have to be made to al- low the necessary analyses to take place. We have found that any artifact, new prod- uct, etc., can be used as a focus to ratio- nalize the formation of the concurrency teams and to guide their activities. RECENT DESIGN STRATEGY EXAMPLES Automobile Factory: Success Based on Technology and Product Redesign Volkswagen's remarkable Hall 54 was re- cently opened to the public. In it, Golfs and Jettas are put through final assembly with 25 percent of the steps done by robots or special machines, as compared with 5 per- cent in the past. The full impact of this change can be
214 better appreciated by considering the con- ventional automotive product cycles. Many products are proposed and undergo devel- opment and prototype design, but only a few are approved. Once a product is ap- proved, financial support for its final devel- opment is assured. At the same time, how- ever, a product introduction date (PID) is set, usually only 24 or 36 months ahead. This date is so near that little time for ra- tionalizing the design is available. Purchase orders for machinery must be negotiated immediately. There is little, if any, time for making a significant change in the design. There is also great reluctance to change the PID unless a major problem arises. To make Hall 54 a success, Volkswagen (VW) obtained approval from its board of directors to delay introducing these cars for a year while "every part was examined" (Hartwich, 1985), and several significant departures from conventional automotive design practices were made. One example is the configuration of the front end. At a cost of adding one extra frame part, the front was temporarily left open so that the engine could be installed by hydraulic arms in one straight upward push. Normally a 1-minute operation or longer, requiring sev- eral men, this process is now accomplished unmanned in 26 seconds. Another example concerns the use of screws with cone-shaped tips. VW introduced these fasteners, which easily go into holes even if the sheet metal or plastic parts are misaligned, at a cost penalty of 18 percent for the screws. This innovation made robot-and-machine inser- tion of screws practical. In the following 2 years, use of cone-shaped screws became so prevalent in Germany that their price has dropped to that of flat-tip screws. Automobile Factory: Success Based on Management New United Motor Manufacturing, Inc. (NUMMI), the joint General Motors (GM)- Toyota company, has been in operation WHITNEY ET AL. since December 1984. It builds a Toyota- designed small car (Ikebuchi, 1986~. The factory has 2,500 team members, 27 Toy- ota managers, and 16 GM managers. The team members were carefully chosen from the group of United Automobile Workers (UAW) that were originally employed at the GM Fremont plant. The UAW, which represents the team members, agreed to many new work rules. In addition, NUMMI undertook a detailed analysis of the main causes for logier productivity and used var- ious techniques to solve the problems that were identified in the following areas: · Quality problems resulted from an as- sumption that repair is a regular process, that standardized work is not practical, that high worker absenteeism must be accepted as normal, that parts and components were regularly damaged during conveyance, and that the low quality of suppliers' parts must be tolerated. · Low line efficiency resulted from inef- fective job classifications and a centrally controlled maintenance group. Low line ef- ficiency created an excessive inventory and a low utilization of equipment. · Excess inventory created difficulties in inventory control, problems with engineer- ing changes, repair problems, too many parts around line workers, and slowness of response time to problems. These problems were solved in the fol- lowing ways: Implementation of ji~ka, the quality principle, in which machines or production stop under abnormal conditions, produc- tion equipment senses malfunction or sub- standard parts, the machine stops itself, and the line workers can stop the line. · Employment of just-in-time UIT) pro- duction and conveyance system that pro- vides only the parts needed, when they are needed, in just the amount they are needed. This requires a kanban system and a system that reduces the die setup time.
THE STRATEGIC APPROACH TO PRODUCT DESIGN · Achievement of high operation avail- ability with foolproof devices, self-monitor- ing devices, light displays, buzzers, electri- cal circuit graphics boards, widely used signs, and simplified job classifications. The basis of the NUMMI system the in- volvement of the production workerscre- ated a yardstick to ensure quality and a guide for employee training. Radiator Factory: Success Based on Integrated Design Nippondenso is the Delco of Japan. It builds generators, alternators, voltage reg- ulators, radiators, antiskid brake systems, and so on. Its main customer is Toyota. Over the years, Nippondenso has learned to be Toyota's supplier, especially how to live with daily orders for thousands of items in an arbitrary model mix. To meet this chal- lenge, Nippondenso has employed several strategies, including in-house development of manufacturing technology, jigless man- ufacturing methods (where possible), and the combinatorial method of meeting model mix. The combinatorial method is the basis of the strategy, so it is described first. A prod- uct is divided into generic parts or subas- semblies, and necessary variations of each are identified. The product is designed so that any combination of types of these basic parts will go together physically and func- tionally. If there are 6 basic parts and 3 varieties of each, then the company can build 36 = 729 different models. The in- house manufacturing team cooperates with the designer of these parts so that the man- ufacturing system can handle each part, possibly by designing common jigging fea- tures onto them, or by advising the design- ers how to make the product hold itself to- gether so that no jigs are needed. The in-house manufacturing team pro- vides other advantages, including protec- tion of proprietary information concerning 215 future product plans or design details, a reduction in overall cost, and a cohesive group that has learned the company's phi- losophy and knows how to contribute. This strategy was employed in the design and manufacture of radiators (Ohta and Hanai, 1986~. As shown in Figure 5, the basic parts are the core (with its basic parts, tubes, and fins), two end plates, and two plastic tanks. Cores and end plates snap to- gether so that they do not need jigs while being oven-soldered together. The tanks are plastic and are crimped on so that prior soldering is not melted. The crimp die can be adjusted between cycles to take any tank size, so radiators can be processed in any model order in any quantity. When asked, "How much did this factory cost?" the chief engineer on this project replied, "Strictly speaking, you have to include the cost of designing the product." WHY IS CONCURRENCY HARD TO IMPLEMENT? Experience has revealed that so many ad- vantages are to be gained from implement- InIe! tank ,~ OEnd pate ~ ELM I A,. Outlet tank End plate Sheet stock ~ Inlet tank Flns ~ Core End plates _ (snap Tubes ~ together) Sheet stock FIGURE 5 Jigless batch-size of one radiator manufacture. Oven Flnlchod: _ ~ Crimp~Test sokler Core ~~' Outlet tank N P9 No 119
216 ing a strong product-process link, one must ask why it is not done more often. A part of the explanation would seem to be as fol- lows: Manufacturing is probably the most complex peacetime activity that people en- gage in. Together with the allied fields of finance and marketing, it can engage hun- dreds or even thousands of people within a single firm. A natural response to complex- ity is specialization, in which people ac- knowledge that they cannot know every- thing. Organizations grow up around specialization boundaries, and people must subscribe to one species or another in order to have a place. By contrast, successful im- plementation of the product-process link re- quires crossing boundaries. Sometimes the ideas that link product and process are too new to be acceptable to established organizations. If the processes are not well understood and seem to require certain "experts" or particular conditions for their success, then conservatism against change may be a rational response. The changes needed in people and orga- nizations to carry out these integrated ef- forts can be difficult to accomplish. But, as the NUMMI example indicates, these changes must be made if the desired pro- ductivity goals are to be achieved. The Na- tional Research Council's Manufacturing Studies Board rightly devoted an entire chapter of its report Toward a New Era in U. S. Manufacturing to this difficult issue (National Research Council, 1986, ~- Chapter 3) . conc-t ~ WHITNEYETAL. enough simply to try harder. A deeper un- derstanding of the fundamental problems is needed. It will require research to identify the new knowledge that is needed as well as to fill the knowledge gaps. Although there are many ways to organize these ques- tions, the scheme indicated in Figures 6 and 7 has been chosen. These figures deal re- spectively with the design of products and the design and operation of manufacturing systems. The issues are set forth in terms of increasing numbers of items (parts or work- stations) and the corresponding increase in complexity. ~ . . . . . Product concept WHAT NEW KNOWLEDGE IS NEEDED? Broad Issues This paper has described a new strategy for improving productivity. The strategy is based on using the assembly process as the focal point and integrator of all the com- plex decisions required to create a produc- ible product. To verify, improve, extend, and implement this strategy, it is not loots that support integrated synthesis, design, and evaluation and data bases to support these tools are needed at each level to help designers and engineers seek alter- natives and to evaluate these alternatives technically and economically. As stated earlier, fabrication by traditional methods has the most advanced design methods, whereas assembly and concurrent design operate with few design tools or none at all. The lack of tools for assembly design is es- pecially acute because the usefulness of as- sembly analysis in forming the total prod- uct design has been recognized only recently, and the outlines of effective strat- egies are only beginning to emerge. In a larger sense, the lack of formal man- ufacturing design tools and computer im- ASPECTS OF PRODUCT DESIGN Separate Into l l _ pare and sub- assernbiles Design Angle parts | _ NEW KNOWLEDGE NEEDED Designer-user Interaction CADICAE and cost modeling New materials Vendor control Functional simulation Design synthesis Design ~ Part mating theory pairs of parts Design ^~l^Q ^' ~?tQ Assembly sequence group. u' pares ~ Subassembly Identincatlon and sub- Tolerancing for assembly sequence assembiles ac strategy Tooling design and accuracy control Detailed simulation Ute-cycle cost models FIGURE 6 Activities and knowledge needed to support the strategic approach to product de- sign: Part 1- Product design.
THE STRATEGIC APPROACH TO PRODUCT DESIGN ASPECTS OF FABRlCAT10N AND ASSEMBLY SYSTEM DESIGN 1 NEW KNOWLEDGE NEEDED Technology choice Fabricatlon | ~ Etilclency station | _ Groups or hbricaUon | swore 1' Fabricatlon systems with transport Technology choice Deafen, byout OpUmizatlon Tool and Nature management 1 Process pan and sequence Match of part types to technology types Operating etilelency Production smoothing Schedule Role of people Economic analysis l Technology choice Assembly ~ Design, byout, optimization stations Error control Groups of assembly stations ~ Technology choice Task assignment Assemb~ wlU,transport Optimum assembly sequence Task a#lgnment Tool change distrlbutlon scheduling for model mix Role of people Economic anolyals FUGUE 7 Activities and knowledge needed to support the strategic approach to product de- sign: Part 2 Design of fabrication and assem- bly systems. plementations of them is a serious defi- ciency for manufacturing engineers. The design engineers have a large array of com- puter and analytical tools at their disposal. This selection gives them a scientific as well as a psychological advantage over manu- facturing engineers. The result is that prod- uct designers are better able to defend their design decisions in the face of attempts by manufacturing engineers to make the de- signs more producible. This result may sat- isfy the product designers but may result in overpriced and noncompetitive products. On the assumption that future products will be designed by teams of product and manufacturing engineers, there is a need for research on the group dynamics of com- plex design projects. Matters of concern in- clude sharing of data, moderating the dom- inance of one constituency over another, defining an effective sequence for making major decisions, and managing the process of negotiating decisions in the presence of conflicting aims. 217 Accompanying the design tools and methodologies must be a large array of data bases. Although many such bases are avail- able for the design of products, more will be needed as new products are created us- ing new materials or processes. Data bases will also be needed for describing the new Processes that will be used. Finally, data bases are needed for the system design methodology itself. For example, there is a need for augmented CAD descriptions of the parts, including notations of possible assembly sequences and call-out of the rel- evant tolerances on jigging and gripping surfaces. There is a need for cost-tolerance data, so that the impact of various process- ing and assembly strategies can be assessed, and for part-mating data linked to the tol- erances, so that overall assembly errors and likelihood of assembly success can be cal- culated. Data bases are needed for fault analysis and fault tree or other methods of representing possible product faults during assembly, so that a quality control strategy can be implemented. Finally, there is a need for data bases on the costs and capa- bilities of manufacturing and assembly methods and equipment that will aid in the economic analysis and synthesis of effective systems. Additional data bases, such as cur- rent warranty data and known field failure modes, may also be relevant. Another important knowledge gap con- cerns performance, including performance models of both the product and the produc- tion systems, and broad aspects of perfor- mance including normal and abnormal op- erating modes, downtimes, repair scenarios, the supporting logistics, and so on. In many cases, the lack of adequate product per- formance models results in overdesigned products. In the case of manufacturing sys- tems, performance models are often nar- rowly drawn with respect to local economic criteria. Too little is known about the cost- benefit relations with respect to flexibility, or how to achieve flexibility by appropri- ately balancing product design, manufac-
218 turing system design, use of new technol- ogy, scheduling and resource allocation, and human effort. Also missing or immature is a strategy for educating engineering students to be effec- tive players in such activities. Interestingly, it appears that Japanese universities are no better than American schools in this regard. Japanese companies, however, appear to compensate for this by typically giving new employees 3 years of training, with rota- tions throughout the company. Our sugges- tion is that students be convinced that a systems approach is the unifying principle, since this will best prepare them for the essential integrative nature of the activity. Once they are aware of the importance of considering many diverse factors before making a design decision, they can then concentrate on becoming expert in one par- ticular area. Specific Near-Term Knowledge Progress is being made in the following specific topical areas: · Methods of generating alternate assem- bly sequences for products. · Algorithms for assessing the tolerances assigned to parts to see if they support a particular assembly sequence. · Procedures for predicting failure modes for a product so that a test strategy can be created for each assembly sequence. · Economic analysis methods for assess- ing (a) the basic assembly cost by various methods, including people and machines; (b) the cost of providing part, tool, and fixture tolerances of different accuracies; and (c) the cost of doing tests at various points in an assembly sequence, including the cost of uncovering the fault by dis- assembly as well as fixing it. The status of each of these topics is dis- cussed in the following sections. Some of the interactions are shown in Figure 8. WHITNEY ET AL. Generating Assembly Sequences The importance of having a good assem- bly sequence has been discussed. To accom- plish this requires that alternatives be gen- erated and then evaluated in some rational way. In the past, alternate assembly se- quences were frequently generated by hand, using either real parts of the product or cutouts of drawings of the parts. This cumbersome process rarely led to a large set of alternatives from which to choose. Tra- ditionally, industrial engineers have used a diagram called a precedence diagram to represent the geometric and other con- straints that limit how a product may be assembled. However, precedence diagrams do not themselves generate assembly se- quences. Furthermore, many real products have constraints that cannot be represented by precedence diagrams. On the basis of prior work (Bourjault, 1984), we have created an algorithm that will enable an engineer to generate all of the physically possible assembly sequences for a product (De Fazio and Whitney, 1987~. The algorithm operates by collating the answers to a series of questions that the engineer asks regarding the assembly op- portunities between related parts. The re- sult is typically many hundreds of se- quences. We also have methods for reducing this number to manageable pro- portions by applying judgment criteria such as how many subassemblies or flip-overs are required. As yet, we have not linked these judgments to economic criteria or to testing strategies, but it is likely that this can be accomplished. Assessing Tolerances In the design of a part, the accuracy of its manufacture must be specified. Some of its surfaces are important to its function, so the designer states tolerances on them for this purpose. However, tolerances must be set for additional surfaces so that assembly
THE STRATEGIC APPROACH TO PRODUCT DESIGN ASSEMBLY SEQUENCE Case Pin Shatt e1 ~ ~ Cover \ act_,, Fully test" subassembly Test-nx costs QC DESIGN TRADE-OFFS - Ta~t too coon Test too late Shatt won't tom Stuck \ Whit ~~Bent ~ Bad I) pin ' ~ assembly - Viable ~ Testable TOLERANCES ~- r' ~ ~ . ~ - JIGGING COST/ACCURACY TRADE-OFFS ~ FAULT TREE Assembly failures and cost ~ Tolerance Too tight Too loose FUGUE 8 The connation be - In the assembly sequence and detailed part design, jig and tool design, and quality control strategy. can take place with confidence that the parts will mate properly. There already ex- ists a large body of theory on how far parts can be misaligned from each other and still be assembled (Whitney, 1982; Whitney et al., 1983~. Thus, the requirement on the designer is to see that the surfaces on which one part rests and the other is grasped are accurately enough made and located on the parts. Naturally, depending on the assem- bly sequence, the resting and grasping sur- faces will be different. This means that the assembly sequence must be known to the parts designer very early in the design pro- cess. By contrast, the usual practice has been to delay consideration of the assembly sequence until after the parts are designed and fabrication methods have been chosen. 219 Since different fabrication methods cost dif- ferent amounts and are capable of making parts to different tolerances, these choices, if made without assembly process knowl- edge, can render an assembly sequence un- realizable. Several kinds of knowledge and design tools are needed. First, we need better models of the costs and accuracy capabili- ties of different fabrication methods. Sec- ond, we need better methods and standards for establishing tolerances. Current meth- ods can be ambiguous, so that a fabricator may not know how the part should be made or inspected. Third, many tolerances com- bine to create the relative locations between two parts. On a typical part, some dimen- sions may have turned out to be at the high
220 end of their allowed range whereas others may be at the low end. If N dimensions or tolerances combine to create the relative lo- cations of two parts about to be mated, then there are 2N candidate worst cases to evaluate. Among these may be some that will prevent the parts from being assem- bled. Algorithms for finding these worst cases efficiently are needed. Linking Failure Modes to Assembly Sequences Every product can experience failures that are due to poorly made parts or prob- lems during assembly. Some of the defects that lead to failures can be detected by in- specting the parts, whereas others do not exist until the parts are mated with other parts, fed electric power or fluid pressure, or pushed or twisted. Thus, the failures be- come critical at certain times during the assembly process, depending on the assem- bly sequence. Furthermore, although a fail- ure may be critical at one point in the as- sembly sequence, it may not be testable until later. Moreover, it may again become untestable still later in the assembly. Fi- nally, the costs of testing and of repair may depend on when the failure is detected, with earlier usually being cheaper. Thus, algorithms are needed that can relate an assembly sequence to a product's fault tree i.e., a diagram of failures and com- binations of failures as well as their mani- festations at test points so that test oppor- tunities, test costs, and rework costs can be assigned to each candidate assembly se- quence. Economic Models of Assembly All of the foregoing design issues have economic consequences that become cart of WHI THEY E T AL. repair strategies. The economic choices pre- sented must be added to those normally made when designs are evaluated and as- sembly methods are chosen. Although we lack the evaluation tools cited, we have al- ready developed other tools that are useful in the design process: · Rework analysis: If the test and repair points and costs of a product are known, and if the failure rates are known, the over- all cost of assembly, including test and re- pair, can be calculated. Such calculations can include products that must be repaired twice or more and can distinguish the need for full disassembly from partial disassem- bly. If a different assembly sequence per- mits lower cost tests, or if a robot has a lower rate of product failure than a person, the cost impact of such alternatives can be calculated. The opportunity to eliminate or substantially reduce reworking has some- times been found to save enough money to justify automation regardless of labor sav- ings (De Fazio, 1986; Gustavson, 1986~. · Unit cost analyses: The cost of assem- bling each product unit can be calculated, given the assembly sequence, the equip- ment or people to be used for assembly, and the usual economic data, such as wage rates, interest rates, production quantity, and taxes (Gustavson, 1983~. Conversely, given a selling price and a profit or return requirement, a maximum budget for an as- sembly system can be determined. · System syntheses: Algorithms are avail- able that can design assembly or fabrication systems (CSDL Reports 1978-1980, R-1284, R-1406, Graves and Lamar, 1983; Holmes, 1987~. That is, given the assembly sequence and data on alternative equipment or peo- ple capable of performing each of the as- sembly steps, the algorithm will select peo- , ple or equipment and assign the steps to the evaluation process. Different designs them, taking into account tool costs, time support different assembly sequences, re- to change tools, and time to transport work quire different tolerances on different sur- from station to station. The selected meth- faces, and require or invite various test and oafs together will produce the product for
THE STRATEGIC APPROACH TO PRODUCT DESIGN 221 minimum unit cost. In a recent study, it will eventually allow a few individuals was shown that different assembly se- working at computer terminals to explore quences for the same product could differ by 5 percent to 20 percent in minimum assembly cost when the algorithm had the same equipment options available (Klein, 1986~. SUMMARY The thrust of this paper is that produc- tive manufacturing systems can be achieved by an integrated approach to manufactur- ing independent of technology, and that the perceived advantages of applying advanced technology can be achieved only through this integrated approach. The integrated approach is more important than any spe- cific technology. A new activity called the strategic approach to product design (SAPD), because of its ability to force com- plicated trade-offs into the open, can act as the catalyst to rationalize product and process design. SAPD is not just a product design method but is also a way of systematizing the way people and manufacturing functions inter- act. It provides a basis for the creation and effective operation of concurrency teams. The method is not an end in itself. Instead, the application of the method provides in- sight into manufacturing systems and their interactions not currently analyzed or un- derstood by any other method. Application of new technology without this integrated approach has proved to be disastrous (Nag, 1986~. Possible Research Initiatives Few individuals have the necessary skills to serve effectively on the concurrency teams. In some industries the average age of those who can operate in this mode is quite high. The nation has a vested interest in capturing their knowledge, creating new knowledge, and developing methods that ~ ~ ~ , new designs with the same range of sensi- tivities that concurrency teams currently muster. The assembly sequence itself is an excel- lent focus for the interaction between prod- uct and manufacturing engineers. The as- sembly sequence can act as the framework for discussions of tolerances, testing strate- gies, assembly methods, and various eco- nomic analyses. A number of analytical and computer tools exist to help in this process, but they are weak except in certain purely economic areas. Several new tools have been identified. A variety of data bases will be needed to support these tools. Such data bases include economic-tolerance-cost mod- els of fabrication methods for parts, fix- tures, and grippers as well as failure mode models of products to permit quality con- trol strategies to be formulated. These new tools would allow logistic and field-support issues, as well as life-cycle costs, to be con- sidered with respect to the total product design and manufacturing system. Education and Institutional Impacts SAPD, concurrency, or advanced manu- facturing system technology requires engi- neers to have a broadly based education. This kind of education is not being offered by U.S. universities at present. Moreover, other areas with the same level of complex- ity, such as large-scale systems, are likewise not being supported. Methods need to be found that allow both the present highly specialized education systems to flourish while encouraging the more broadly based training. A number of universities, with the support of the Na- tional Science Foundation and the Depart- ment of Defense, are exploring the role of centers of excellence to carry out this mis- sion. It is not clear whether the complexity of advanced manufacturing can be ade- quately explored with this method.
222 Most of the manufacturing system design decisions needed, as well as the skills of the people who make these decisions, are cur- rently based on experience. As industry moves from this experience base of opera- tion to a science base, severe displacements in the ranks of middle management can be expected. Chapter 3 of the Manufacturing Studies Board report (National Research Council, 1986) explores several facets of concern and suggests that a different kind of organization will result, one more orga- n~zed as a team than the conventional'ad- versarial management-worker relationship. In such a team, all members will partici- pate in the decisions. The NUMMI experi- ence certainly supports this view. ACKNOWLEDGMENT Many of the ideas in this paper appeared in "What Progressive Companies Are Do- ing to Raise Productivity," by J. L. Sevens and D. E. Whitney, prepared for the De- fense Manufacturing Forum, "Rethinking DoD Manufacturing Improvement Strate- g~es," at the Institute for Defense Analyses, October 29, 1986. REF ERENCES Akagi, S., R. Yokoyama, and K. Ito. 1984. Optimal design of semisubmersible's form based on systems analysis. ASME Paper 84-DET-87, ASME Journal on Mechanisms, Transmissions, and Automation in Design 106(4):524-530. Boothroyd, G., C. Poll, and L. Murck. 1982. Auto- matic Assembly. New York: Marcel Dekker. Bourjault, A. 1984. Contribution a Une Approche Methodologique de ['Assemblage Automatise: Elab- oration Automatique des Sequences Operatoires. Ph.D. dissertation, Universite de Franche-Comte. Chirillo, L. D. 1982. Product Work Breakdown Structure. Maritime Administration, National Shipbuilding Research Program, U.S. Department of Transportation. CSDL Reports. 1979-1982: R-1407, R-1537. 1974- 1980: R-800, R-850, R-921, R-996, R-1111, R-1276. 1978-1980: R-1284, R-1406. Charles Stark Draper Laboratory, Cambridge, Mass. De Fazio, T. L. 1986. Uncertainty in Unit Costs Oc- WHITNEY ET AL. curring During Low-Throughput Operation of Process Systems That Include Testing and Rework. MAT Memo 1299. Charles Stark Draper Labora- tory, Cambridge, Mass. De Fazio, T. L.7 and D. E. Whitney. 1987. Simplified generation of all assembly sequences. IEEE Journal on Robotics and Automation. In press. Graves, S. C., and B. W. Lamar. 1983. An integer programming procedure for assembly system design problems. Operations Research 31 (3) :522. Gustavson, R. E. 1983. Choosing manufacturing sys- tems based on unit cost. Pp. 85-104 in Proceedings of the 13th International Symposium on Industrial Robots, Chicago. Dearborn, Mich.: Society of Manufacturing Engineers. Gustavson, R. E. 1986. A Statistical Analysis of Re- cycling, Rework, Yield, and Cost Reduction. MAT Memo 1300. Charles Stark Draper Laboratory: Cambridge, Mass. Hartwich, E. H. 1985. Possibilities and trends for the application of automated handling and assembly systems in the automotive industry. Pp. 126-131 in International Congress for Metalworking and Au- tomation, 6th EMO, Hanover, FRG. Holmes, C. A. 1987. Equipment Selection and Task Assignment for Multiproduct Assembly System De- sign. S.M. thesis. Massachusetts Institute of Tech- nology Operations Research Center, Cambridge. Ikebuchi, K. 1986. Unpublished speech. Conference on Future Role of Automated Manufacturing, New York University. Klein, C. J. 1986. Generation and Evaluation of As- sembly Sequence Alternatives. S.M. thesis. Massa- chusetts Institute of Technology Mechanical Engi- neering Department, Cambridge. Nag, A. 1986. Tricky technology: Auto makers dis- cover factory of the future is headache just now. Wall Street Journal 13 May: 1. National Research Council (NRC). 1986. Toward a New Era in U.S. Manufacturing: The Need for a National Vision. Manufacturing Studies Board, Commission on Engineering and Technical Sys- tems. Washington, D.C.: National Academy Press. Ohta, K., and M. Hanai. 1986. Flexible automated production system for automotive radiators. Pp. 553-558 in Proceedings of the First Japan-USA Symposium on Flexible Automation, Osaka, Japan. Olmer, L. H. 1985. U.S. manufacturing at a cross- roads: Surviving and prospering in a more compet- itive global economy. International Trade Admin- istration, U.S. Department of Commerce. Opitz, H. 1967. A Classification System to Describe Workpieces. Oxford, England: Pergamon Press. Porter, M. E. 1986. Why U.S. business is falling be- hind; the country is investing too little in the tech- nology, facilities, and education it needs in today's market. Fortune 113(April 28):255-262.
THE STRATEGIC APPROACH TO PRODUCT DESIGN Taylor, F. W. 1911. The Principles of Scientific Man- agement. New York and London: Harper and Brothers. Wheelwright, S. C., and R. H. Hayes. 1985. Com- peting through manufacturing. Harvard Business Review 63~1):99-109. Whitney, D. E. 1982. Quasi-Static Assembly of Com- pliantly Supported Rigid Parts. ASME Journal of Dynamic Systems, Measurement, and Control 104:65. 223 Whitney, D. E., R. E. Gustavson, and M. P. Hennes- sey. 1983. Designing chamfers. International Jour- nal of Robotics Research 2:3-18. Whitney, D. E., T. De Fazio, R. Gustavson, J. Nev- ins, D. Seizer, and T. Stepien. 1986. Implementa- tion Plan for Flexible Automation in U.S. Ship- yards. CSDL report prepared for Todd Pacific Shipyard, Los Angeles Division, and SNAME Ship Production Committee Panel SP-10, Flexible Auto- mation.
