Integrated Design and Manufacturing in Japan
The Japanese approach to product creation rests on recognition of the strategic importance of the overall product design capability
Daniel E. Whitney
In 1991 I spent four months in Japan visiting companies enterprise process of determining customer needs and con to study how they use computers in the design verting them to concepts, detailed designs, process plans, Among the of their products. My goals were to determine the outlines factory designs, and delivered products, together with advantages of of the product development process; to find out what their supporting services. This has come to be called the shortening computer tools are in use and where they come from; to product creation process (PCP). learn how the needs of marketing, manufacturing, leadtime are that The Japanese Approach to Product Creation designers learn to assembly, field service, and other areas are taken into The largest and most advanced of the companies I account during design; and to ask what the Japanese see visited generally take a total-enterprise approach to design. do design better as the main blockages to better product design. In my terms, they are vertically integrated in the skills and engineers get I made 25 visits to 15 companies, which range in size and technologies of product realization. One of the most more practice at from 1700 to 450,000 employees and represent heavy, important findings was that Japanese firms - unlike engineering. medium, and precision industries. Several of these firms American or European firms - recognize the strategic are in more than one industry segment, and in several importance of creating both their own manufacturing cases I observed the same product at more than one com eqUipment and their own design software. Similarly, they pany (such as cars, video cameras). Visits typically lasted recognize the central importance of their design capabili one day and consisted of interviews with designers and ty to their overall way of doing business. The main priori their managers, plus visits to design and prototype labora ty they cited was to shorten the leadtime for developing tories and computer rooms. In terms of their approaches new products and thereby to shorten the cycle from one to product design, their priorities, and the methods they product generation to the next. Among the advantages of use, I found remarkable uniformity among them. shortening leadtime are that designers learn to do design By product design I mean more than how a product better and engineers get more practice at engineering. looks and feels to its user. My findings comprise brake These improvements compound, adding up to a growing systems, machine tools, and other aspects of products for lead. Other priorities include shortening the response which appearance is far less important than engineering time to customer orders, standardiZing design practices performance. Design in this context means the total- worldwide, deploying proprietary databases and design
7 March/April 1994 rules, connecting all the designers to a common data components company, indicates early release of partial base, and enabling unique manufacturing methods that design information. When I asked at Toyota if this was an give a competitive advantage. informal release "between friends," I was told emphatical I found that the principal management method ly, "No." Every release is approved by some manager, but Human being used is task overlapping, that is, beginning to official releases have approval from higher management. experience design the manufacturing system and equipment while The trailing tail is called "continuing responsibility" at dominates the product is still being designed. This approach, com Toyota, meaning that the designer continues to be calculations and monly called "concurrent engineering" in Japan, far responsible for the design even after it leaves his hands. If exceeds the typical U.S. definition of concurrent engi changes need to be made later to help manufacturing or formal data; neering: that the manufacturing people give plenty of another function, the original designer makes them, and human feedback to the designers so that the product can be made thus closing the information loop directly. At communication, efficiently.! The Japanese have been doing the latter for Nippondenso, the leading tail includes "early warning," mostlyface-toface, years and frankly wonder what U.S. companies mean by meaning that downstream people prOVide critiques and is the main tool for "concurrent engineering." updates on the suitability of designs to the upstream The Japanese meaning of "concurrent engineering" people. Nippondenso does not specifically identify a linking design trailing tail activity. activities. is shown in Figure 1. Drawn by a Toyota design manager, this diagram breaks up the car design process into The Human Element approximate generic steps and shows where the official Another important finding was that the bulk of release of information occurs from one step to the next. Japanese design effort is, in fact, human. The penetra The height of each balloon schematically represents the tion of computers is very uneven, though growing in manpower loading of a design step over time. The lead focused and fascinating ways. Human experience domi ing tail of each balloon, called "flying start" at Toyota nates calculations and formal data; and human com and "front loading" at Nippondenso, an automotive munication, mostly face-to-face, is the main tool for linking design activities. My hosts boasted of the strength of their engineers' Flying Start and Continuing Responsibility "universal experience," gained in on-the-job training. They feared electronic concurrent engineering based on Market Concept electronic mail, expreSSing concern that it would impair Official . face-to-face communication. On-the-job training is ;SfOf"'~ release required because new Japanese engineering graduates information ~ ..Styling have surprisingly little specific, deep knowledge in any .... Continuing responsibility one engineering field. University curricula do not contain F/ying...... :4l>- Preliminary design reqUired subjects, and many classes resemble survey Start courses. Students thus graduate with wide exposure to many technologies but without much deep knowledge or ~ Detail design commitment to a particular technical discipline. Though American graduates are trained far more deeply, their o Prototypes backgrounds are narrower. Whereas U.S. companies com plain about American students' lack of preparation for o Evaluation the real world and see the necessary in-house training characteristically as a burden, Japanese companies rec ~ Tooling design ognize that schools cannot provide real-world experience and look on in-house training as an opportunity. -
10 Target Volume 10, Number 2 Ultimate Cost and Quality - Product Design Product Design
Ultimate coyt Typical focus of and to me, "We used to buy computer systems and adapt our concurrent Quality work style to suit. Now we are formulating our next-gener engineering ation work style and will write or buy software to suit." Facility and Equipment This person, like his counterparts at other Japanese com Design/Build panies, is a former engineer - not from MIS - and he knows the situation and trends better than either of the Figure 2. professors. To him, it is not a matter of management methods versus computers but a combination of both that must be tailored to a company's methods and markets process and are often called CAD planning departments. and to the particular technologies of its products. He is Their job proceeds in two steps. The first is to create what saying three very important things: are commonly called "islands of automation" in the manufacturing arena; the name applies equally in the • His company recognizes something called a design arena. An engineer or member of a CAD planning "working style." department writes a program to do a specific calculation, • The company pays very close attention to it, assigning such as a stress analysis of a connecting rod or fluid flow a separate team to oversee it and subject it to continu in an intake manifold. Repeated use of the program ous improvement. improves its accuracy, and perhaps other programs are • Working style is tightly linked to, and sets the require written to do other calculations. ments for, design software. This process may continue, with the islands getting Closely related to in-house design of software is inter better and better. Alternatively, someone may decide that the time has come to link these programs together so that nal design and construction of facilities. In Figure 2, the all the calculations for design and manufacture of the two elements across the top are most often cited in cur item can be done without manually transferring the out rent U.S. literature on improving design and manufac put of one program into the input of the next. The obvi turing. According to the Japanese company that offered ous question is whether the individual islands are me this diagram, the picture is not complete until the accurate or mature enough to justify the large effort of third leg is added. To support all three legs, this company integration, or whether some steps required for a fully has written its own CAD software, several additional pro integrated system are still unsupported except by primi grams that calculate important engineering performance tive estimates, experience, or other manual methods. gauges, and its own design-for-assembly software, which I have found that several Japanese companies have it uses in a unique way as discussed below. identified this integration step as distinct and worth The role of computers in the PCP is governed by long pursuing even if the islands are unevenly developed. experience with a manual design process. Anumber of You can't There seem to be two reasons: First, integration brings my hosts contrasted the U.S. tendency to computerize great speed advantages, eliminates manual data trans automate quickly without understanding the process, thus auto fer errors, and puts many dispersed decisions in the something you mating some very inefficient procedures. At the same hands of one or a few people, whose expertise thereby don't understand. time, they criticized themselves for being too slow to rises. Manual intervention, such as correcting oversim computerize, always looking for the next improvement in plified computer results, may be used to bridge over the the process. Yet their logic is compelling and has a strong weak spots. parallel in traditional process automation: You can't Second, integration represents a learning opportunity automate something you don't understand. and challenge of its own, which can be carried on concur From Islands to Systems rently with longer-term efforts to perfect the islands. The Japanese technique is to develop and learn the Integration requires marrying previously separate data, process manually, subject it to continuous improvement, methods, and people, and reveals previously undetected conduct postmortems, and establish separate organiza errors in data and inconsistencies in methods. It also tions charged with identifying opportunities to computer tends to connect and even equalize functions that are ize. These organizations are staffed by veterans of the often culturally separated, such as drafting and engineer- 11 March/April 1994 ing calculations, creating an engineering-driven process in Figure 3 allows the stylist to continue using studio rather than a draWing-driven process. In this way, skills and methods in the new software environment. CAD changes character entirely. Computerized integration has another powerful The most vivid example is Toyota's integrated car advantage: It permits rapid changes in the design and body styling, body engineering, body stamping die allows those changes to ripple efficiently through the pro design, and die manufacture (see Figure 3). This is an cess. I noted that the overlapping tasks method is risky. The unbroken digital design, data, engineering, and fabrica risk, of course, is that later tasks will use information tion chain from the input of stylists' sketches to the that is incomplete or imprecise and will have to be par machining of the dies. The process includes a crucial step tially done over if that information changes. I asked my called "formability analysis," which determines whether hosts about this and was told repeatedly that change is the stylists' shapes can be made or not. The analysis must inevitable. "Much larger changes come from watching be done early and qUickly so that the outer body data can the market and our competitors. Changes caused by task be frozen and follow-on design can occur. Time-consum overlapping are usually less severe." The message seems ing clay models are not used to create design concepts, to be that, rather than fighting change, these companies but only as displays; they are carved by computer have decided to learn how to live with it. Integrated com controlled milling machines directly from the database. puter tools are one way of doing so. At Toyota, body styling and engineering is almost The design software these japanese companies are At Toyota! body paperless. The stylist sketches original designs as before, writing is mostly advanced three-dimensional solid or sur styling and but these are entered into the computer by a skilled com face modeling, although some companies are still using engineering is puter graphics operator. These two people then refine the two-dimensional drafting software they wrote years ago. almostpaperless. design together. They use simulations of many common While Toyota and Nissan both date their first efforts in design studio tools. For example, they shine simulated CAD to around 1965 (when Ford and GM also made their lights on the shape in order to judge the reflections. The first efforts), the software I was shown appears to have final shape is carved into a quarter-scale clay model for been first written in the late 1970s, with many upgrades executive review. Such a cycle takes only a few weeks, since. The smallest company I found to have written such compared to several months by the old clay model advanced software has 13,000 employees; smaller compa method. The resulting data are used to launch body engi nies had bought nearly everything. The largest companies neering, formability analyses, design of inner support have integrated some key design and analysis sequences, panels, creation of stamping dies and associated tooling, such as the Toyota approach discussed above. generation of manufacturing schedules, and creation of Whether or not they have the resources to produce cutting machine and inspection machine commands. their own design software, japanese companies imple Toyota has been using this system for about five ment massively whatever they decide to use. In japan, years. While it is not without shortcomings, it's the best of engineers do their own drawing; there are no draftsmen. its kind. Of course, every car company does the same cal Every company I visited had its engineers working on culations in its own way, and many of these are comput networked computer terminals, not drawing boards. One erized. But none are as fully integrated. Though Toyota's computer or terminal for every two engineers is judged advanced formability analyses are admittedly simple and sufficient, though most companies have fewer computers based on pre-existing hand-calculation methods, they at the moment. It was impressive to walk into one auto are familiar to the engineers, avoid all known disasters, motive company's design center and be told that there were and ensure that die design, always a time-consuming 1400 design terminals and 3000 PCs (including 1000 and expensive process, is shorter because much of the Apple Macintosh computers) all networked together. trial and error has been eliminated. The simplicity of the analyses ensures that the engineers understand and trust Computer-Aided Engineering them - an essential ingredient in getting the engineers Most companies buy their generic computer-aided to use the system. engineer (CAE) software, principally from the United Careful attention to the user interface is also essential. States. This includes stress, thermal, fluid flow, and car For example, the item labeled "highlight line simulation" crash simulations (the latter from France), analyses of 12 Target Volume 10, Number 2 Die Design to Fabrication Integration
Style CAD Data Clay Model B d CAD D t (NC Milling) 0 y a a ~ Sketch ~
Press Die CAD Data VfasterO,ata for Manufacturing Functions for High Quality Surfaces :expansion~ ' die ~ ~-.f ~~ Smoothing ~anel ..kim line ~ pI] Curve fitting Formability Expectation "~W fillet : over crown Highlight line .' expansion simulation ~ db~~'" /' ~~~Mecting ~ Fairing ~ Linking curves die face' surfaces Fillet Modeling Full Mold Model _--_ (NC Milling) Cutler ~~:~:Yledge y Location Surface modeling ::. Data
~o ~Q t:::i with Automatic Calculation a Cutting of Cutter Location Conditions Casting•Work NC Milling ~~ Complex Clearance for ~ surfaces the good fitness Another Variation of Cutting ~~ Profile Character line Measuring and Analysis punChn --Jf\ NC Scribing , y~ Rtouching 3D Measuring Machine ~ ~ point line
Figure 3. Toyota's integrated approach to the design-to-die manufacturing cycle takes months out of traditional clay model methods. Toyota has been using this approach for five years. Source of information: Toyota.
13 March/April 1994 how molds and castings fill and cool (from Australia), it. Instead, assembly is considered more or less continu how robots and machine tools move, and so on. ously throughout the process. Many Japanese companies have written specific One company I visited must deal with an enormous engineering analysis programs. Examples include opti number of versions of its products. Its DFA process gives mizing internal shapes of dot matrix print heads so that preference to parts and assemblies that can be switched the print wires can be inserted by robot; tracing rays of easily from one version to another or that are common light through multiple lenses, accounting for a camera to many versions. Another company makes intricate company's wide range of proprietary optical glasses; fol mechanical devices to be assembled by robots; product lowing individual fuel particles from the injector exits to life is about 18 months. Assembly analysis has been valve openings so that gas mileage can be raised; and compressed into the concept design phase, and DFA improving strength-to-weight ratio of mechanical parts scores are used in selecting concepts for further develop so that products will be lighter but won't rattle or ment. Another company classifies parts and assemblies vibrate, even though they are thinner or more slender. into "critical" and "secondary," then attempts to sim Except in the areas of car body styling and engineer plify the critical ones and eliminate the secondary ones. ing, there is less integration of CAE than there is of con Secondary ones that present assembly difficulty (a float ventional design and manufacturing. CAE still consists ing attached wire, for example) get especially low of islands of automation with awkward conversions of scores. In the United States, DFA at its best is used to data from the design database to the form needed for simplify the design by consolidating parts. However, it is analysis. On the other hand, integrated conventional often delayed until the details of the design are known, design and manufacturing, without any intervening sometimes even until prototype parts are available. analysis, is much more common. Atypical example is The most limited scope for DFA is to make assembly the creation of three-dimensional part shapes in the easier. The broadest scope I found in Japan was at computer, followed by automatic process planning for Nippondenso. Here, DFA has been merged with Design manufacturing them and automatic generation of for Manufacture (DFM). The intent is not just to simpli numerical control commands to the machine tools. fy manufacturing and assembly, but to enable entirely Such applications can be found at many advanced new methods of making a product. New methods are companies worldwide. However, the number of small required because Nippondenso faces a relentless companies in Japan that are at this stage may be unusu increase in the variety of its product models. Its products al. This fact may reflect the strength of the supplier base are sold mostly to OEMs and face stiff cost-performance underlying the large companies. competition. At the same time, because its customers Design for Assembly practice just-in-time manufacture, ways must be found Attention to While automation of the design and fabrication of to switch models at arbitrary intervals without any assembly is an single parts has received a great deal of attention for "cost" in downtime, investment in fixtures, or excess indicator- of decades, assembly has only recently been recognized as capacity. To meet the high production volumes and an important area for design improvement. Thus, atten high quality standards of its customers, Nippondenso maturity in a tion to assembly is an indicator of maturity in a compa automates as much of its assembly as possible. But company's PCP. ny's PCP. One-third of my hosts have written their own automation normally does not support the rapid Design for Assembly (DFA) software. All have taken dif changeovers that are needed. Nippondenso identified ferent and interesting approaches, either in defining this basic challenge nearly 20 years ago and has had an what to judge about an assembly or in choosing how to ongoing program to meet it ever since. use DFA to influence the design process as awhole. What Over this period, the company has solved successive I saw was at least as sophisticated as the DFA practiced ly harder variety-changeover-automation problems. in the United States. Some companies have raised it to First, it attacked small items, using a method in which an art. Others wonder what the fuss is about and have parts of nearly the same shape and size but different not identified a specific DFA step in their PCP or con function were substituted for each other to create the ceived specific procedures or software for accomplishing required variations. It then extended this method to
14 Target Volume 10, Number 2 large products. Most recently, the method has been write design software for in-house use and for sale to applied to products in which different models have radi their respective suppliers in order to unify the design cally different sizes. This is the hardest kind of variety to process and data formats up and down the supplier absorb on one piece of automated equipment; the typi chain. PC versions are available to smaller suppliers. cal response is to build one machine for each size or to For these companies, the advantages of "doing it narrow the size range. Since demand for each size is so yourself" clearly outweigh the risks. These advantages unpredictable, the usual method results in overcapacity, include the obvious ones, such as design speed, ease wasting money and floor space, the latter being critical of sharing data, and accuracy of results that come from ly short in japan. integration. Less obvious advantages include the ability Risks and Advantages to absorb change more e.asily; to accumulate and Companies that commit to making key manufac retrieve corporate design data; to capture a home-grown turing equipment shoulder the responsibility for creat and carefully tuned design methodology, standardize it, ing and maintaining a wide variety of capabilities and and make the designers use it; to empower new hires to skills. As with other things, this responsibility is either a operate at nearly the level of experienced engineers; to burden or an opportunity. At one U.S. company that transfer japanese methods to U.S. and European rejected it as a burden, a somewhat dissident engineer employees as the companies expand worldwide; to told me, "When we have a problem and do a root cause transfer design skills and facilities to members of the analysis, one of the causes often is 'we do not have con supply chain Keiretsu as part of the larger process of trol over design and production of the manufacturing improving suppliers' capabilities (a major strategy of equipment.' Yet top management does not see the japanese supplier relations); to gain control of the pro importance of this lack of control." On the other hand, cess of continually imprOVing the PCP; and to decouple at a japanese car company known for buying non the creation of advanced design methods from the strategic components (that is, components not in the somewhat limited capabilities of CAD vendors who want crucial body-drive-train-interior group), I saw equip to sell generic products. ment being built to manufacture a simple commodity Long-Term Trends suspension strut that could easily have been bought. The boundary between design and engineering is Apparently, building a non-strategic part and the neces blurring. Design is crossing cultures and time zones, sary manufacturing equipment in-house is deemed a The boundary playing a key role in a company's ability to create prod small price to pay for having the skills available when between design they are really needed for building strategic parts. ucts and production methods mated to customers and and engineering Similarly, a well-known risk in "rolling your own," markets. Old-style geometric modeling is now less is blurring. as in-house software development is called, is that the important than the management, organization, company must keep up the software, support training retrieval, and processing of all the information, includ and user manuals, and absorb changes in computer ing geometry, that characterizes modern design. technology. The japanese are now facing the last of The capabilities needed to support these trends are these tasks with some urgency, because most of their not available as off-the-shelf products from the CAD integrated CAD/CAM/CAE is based on large mainframe industry. Acompany that wants a distinctive and superi computers that can no longer prOVide the speed of or PCP must take responsibility for first developing its response that more complex product designs and mod own and then supporting it from within. eling software demand. Conversion to workstations is on Once, one person could hold in his or her head all their agenda, but this technology is changing so rapidly the knowledge needed to design and manufacture a that no one is eager to make the switch. Even so, Toyota musket, a bicycle, or even an automobile. When prod has 350 outside programmers plus 200 of its own con ucts became too complex for this method, the "Ford stantly writing and supporting a wide variety of software approach" was applied, in which designers and engi in-house. Both Toyota and Nissan have long-standing neers were specialized. Each did his or her little part of strategic alliances with IBM japan and Nihon Unisys to the job and passed it on, along with the responsibility, to
15 March/April 1994 the next person. The disadvantages of this approach Daniel E. Whitney is with the Design and Automation Technology were eventually recognized, and assembly lines gave Group at the Charles Stark Draper Laboratory in Cambridge, way to the cross-functional team approach. Massachusetts. His work includes research and consulting in As we head into the next century, many U.S. compa product design, design methodologies and best practices, and nies are using the team approach - while in]apan the development ofcomputerized design tools. He wishes to thank the Charles Stark Draper Laboratory, Inc., for its support during this next-generation approach is emerging. There, the indi study; the Office of Naval Research Washington and Tokyo vidual steps in the team approach, including the data offices; and the staff of the companies and universities who requirements and flows, the calculations and displays, served as his hosts. are being studied in detail. Places where the process depends on increasingly scarce human experience are being targeted for extra data gathering or artificial intelligence approaches. The arrangement and structure of the PCP are being optimized, and key portions are being transferred to computers. The most advanced companies realize that a computerized PCP is not a commodity, but a proprietary resource that captures internal knowledge and data that were developed from and depend on the technical characteristics of their products and the experience of their people. This kind of PCp, almost by definition, cannot be bought, but must © 1994AME® For information on reprints, contact: be built in-house. Association for Manufacturing Excellence 380 West Palatine Road 1. See Daniel Whitney and Michael Nevins, Concurrent Wheeling, Illinois 60090-5863 Engineering, McGraw-Hill, New York, 1992. 7081520-3282
16 Target Volume 10, Number 2