Therefore, it is now accepted that decisions made during design dominate the cost and performance of the product, and most downstream decisions are subordinate, with little ability to affect cost or performance. Thus intense attention is being given to how to accomplish CD. Even Toyota and Honda, according to Fujimoto, feel the need to do this better.

Honda used to have a separate design "company" called Honda R&D, which essentially sold finished car designs to Honda Motors. The manufacturing system engineers at Honda Motors prided themselves on their flexibility, their ability to accept a lot of "wild changes at the last minute," though not after the product was launched. My own opinion is that in the United States the wish of manufacturing engineers to appear flexible has been a barrier to increased CD. Until the early 1980s in the United States, the quickest way to be swept aside was to complain that the design needed to be changed. It is possible that Honda R&D's designers knew enough about producibility that they could avoid a lot of obvious problems. U.S. designers are less likely to have such knowledge.

Particular Fujimoto/Clark Findings

In automobile companies, the time to design and field a car model is dominated by the time to design the body and the stamping dies needed to make the body parts. This time is, in turn, dominated by the time to design, make, and test the dies themselves. In Japanese companies, the lead time for dies is about 1 year versus 2 years in the United States and Europe. Furthermore, the impact of a die design change on both cost and schedule is much lower (10% to 20% excess cost over budget versus 30% to 50% in the United States and Europe). On the other hand, the absolute number of engineering changes is not too different, he says (500/month

versus 1,000/month). Thus the difference must lie in the extent of the changes: apparently they can be rather large and disruptive in the United States whereas in Japan communication has ironed out the big problems early and only small ones crop up.

The difference between regional performance here lies in the organization of the design process, the timing of information transfer between phases, and the type of information transferred. Japanese production tooling designers can use sketches, line drawings, or preliminary drawings to begin their own designs. In the United States, tooling designers are reluctant to use such information and body designers are reluctant to provide it because changes give rise to blame-fixing and fingerpointing. Furthermore, some information is not directly obtained in Japan but rather is intuited in the sense that designers know each other and their styles, as well as the most change-prone regions of a design. Thus they can anticipate some changes by providing safety factors, such as excess metal on the die face that will be removed later when final designs are in hand.

Strangely, much of the information of most forward use to die designers exists when the final clay models are finished, which is before any body panel design begins. According to Fujimoto, the die designers "lack access" to the clays, which can be many miles away, or locked behind elaborate security barriers. Another source of such information that is not used is the model shops where preliminary dies are designed. Habit plays a big role in such lack of information transfer.

Die designers are not used to being asked what information they would need and thus have not taken the time to decide what would be of the most use. They take it all in at once without partitioning it in their minds or in their design process. Formal methods to partition a design process are only recently drawing attention as a research

topic and may indeed represent a new way of thinking (Ref 5). On the other hand, Japanese designers, when forced to do things faster, may have thought the problem through and made priority lists of what information they really need early in the die design process. They may have also found out that such information may be easier to obtain than they thought, especially if they promise not to complain if the design changes later. In fact, Fujimoto claims that overlapping of the process steps came first, as a means to shorten the cycle, and communication lines developed in response. (See the Toyota reports, where the claim is made that information release is tightly controlled and does not occur freely between longtime acquaintances.)

As another example, in the aircraft engine business, it has been found that the most useful advance information for tooling engineers about an engine shaft is its length and outer diameter. This information permits them to order raw material and begin machine design. However, the shaft's inner diameter does not strongly influence preliminary machine design and can be learned later. Also, preliminary finite element method (FEM) analysis can be done knowing only a rough inner diameter, and this analysis helps determine what the final inner diameter must be.

Another important factor is a company's reward structure and management culture. Companies that penalize design changes may stifle early communication and overlapping of processes and force designers to furnish only "final" designs.

Human Communication Versus Computer Implementations

Fujimoto/Clark represent a research paradigm that is not surprising in a business school, namely, that management techniques and human activities dominate, and computer tools and other

advanced design and manufacturing technology are subordinate. Toyota seems to be living proof of this. Fujimoto extends this claim by pointing out that many aspects of car design simply cannot be handled long distance by any existing technology, such as electronic mail or real time links between computer workstations. "Only face to face communication will suffice." Examples concern how to capture certain human expressions in the design, such as the feel of the suspension or the handling, or even the sound the engine and exhaust system will make (so-called esthetics of engines mentioned above). Fujimoto/ Clark posit the existence of "heavyweight product design managers" as the main carriers of this type of information (Ref 6). Such managers are deeply experienced in both market and engineering aspects of car design and have been raised from a background in engineering. Their use has been adopted in the car industry world-wide as car designs have become more sophisticated and complex. Heavyweights do not simply push schedule and budget but meet with designers every day, personally drive the cars on the test track, talk intensely with the test engineers, and so on.

However, an alternate paradigm exists in the engineering schools and in engineering research on the design process. This paradigm states that advanced computer tools are essential to the design of complex products, the prime examples being microprocessors, aircraft, and cars. The unification of design data, production orders, market information, and control commands to production machines is a major goal of modern integrated manufacturing and is the subject of the emerging IMS project (Intelligent Manufacturing System) among others. Another similar effort is the international standards project called PDES (Product Data Exchange Standard using STEP). (STEP is the European acronym for PDES.) PDES will establish a data exchange standard

that will permit all the data describing a product to be converted and communicated within and between factories, design studios, suppliers, vendors, and

so on.

Fujimoto argues that communication capability is an organizational skill that takes years to build up. It is based on personal contacts developed over years of working on similar design projects and is fortified by the stability of employment and length of job assignments typical of Japanese companies. By contrast, he views computers as commodities that anyone can buy and install quickly. Thus no competitive advantage is gained by going this route, although anyone who does not will be left behind. But adopting the right communication methods is not a commodity and others cannot quickly jump in and achieve the same results.

The engineer paradigm people view the situation exactly oppositely. There is increasing evidence in U.S. companies that communication and overlapping of product design and production equipment design can be introduced quickly and effectively. Methods used so far require intense personal pressure and long working hours (same as in Japan!) and considerable fatigue results. This means that such methods have not been institutionalized yet, but they will be, at least in the survivor companies.

However, computers may not be commodities after all. It is true that workstations are becoming commodities, and so is basic commercial design software, such as CAD tools. However, the true differences between companies will soon shift: right now the difference is in how much CAD has been adopted. In the future, the difference will be how much beyond CAD is achieved. This includes capturing design knowledge from past designs in computer-readable form, making product design data available to tooling designers, creating computer “design critics" that can automatically comment on a product design and alert designers to producibility

problems, integrating engineering and business data, and so on. Much of this kind of information is currently embodied in the experience of each company's engineers, meaning that new kinds of software tools and knowledge capture methods will be needed. Companies that put effort into this kind of computing will definitely gain a competitive advantage that others will not be easily able to copy.

The willingness of companies to invest internally in computerized design infrastructures, including supporting research on advanced design tools and methodologies, may be an essential factor in future productivity and commercial health. This will be especially important as cadres of experienced engineers retire and take their experience with them.

Final Comment

Clark and Fujimoto have focussed on one industry, automobiles, one aspect of car design, some moderately complex parts with aesthetic elements, and one period of industrial history, the 1980s. Other industries are characterized differently, as are other items they make. Electronic component design is heavily computerized and data driven, whereas almost no mechanical design is similarly structured. In another report, I describe a Japanese company's methods and equipment for designing small VCR mechanisms as well as their outlook for future competitive advantage. While cars are designed by teams numbering in the thousands, these VCRs are designed by a team of 10 engineers total. Communication problems clearly vary with the scale of the project. Yet, with 300 parts, VCR mechanisms are not simple.

Fujimoto added in later a communication with me that the characteristics of the auto industry in the 1980s may not apply to the 1990s. Indeed, computers and their constituent software and databases may become the

discussed the product design cycle for miniature VCRs, including how long it takes, what the chief obstacles are, and who does what. I was also shown the CAD center.

After lunch, I got a brief tour of PERL's robotics laboratory where I saw several things I had seen in earlier stages of development on previous visits (six in all since 1974).

Finally, there was a presentation by PERL of a recent paper on their approach to Concurrent Design (see the report on my visit to Prof. Fujimoto) and an open discussion on this topic. We planned several additional visits, including to factories where washing machines and automotive components are made. These visits are covered in other reports.

VCR Design Cycle

I&MSL is responsible for consumer stand-alone VCRs, VCRs inside camcorders, and professional studio VCRs. The latter are their newest products and are made at the glacial rate of 10 per month. By contrast, the others are made in the hundreds of thousands per month. Naturally, the design of something made at 10 per month is quite different from 100,000 per month, since automation of the latter is almost a necessity. Thus each part must be very simple, must be able to be picked up by a simple gripper, and installed in a simple straight-line motion.

These VCRs are an incredible jumble of stacked, nested, and intertwined levers, rollers, springs, sliders, cams, and bars, run by a motor and several little rubber belts. Most of the parts occupy the bottom 3-mm-thick layer of a unit that is about 2 cm thick, 10 by 14 cm. Design is therefore an exacting process requiring fitting parts into small spaces as well as determining how the moving parts will travel while carrying the tape from the cassette to the read head.

A totally new product may absorb 2 years of advanced development of its basic technologies before any specific product design begins. Then it takes about 2 more years to create a product. For a relatively mature product, one of these years will be devoted to design development at the laboratory while the other year will be spent converting it into a manufacturable item utilizing the works designers. For less mature products, correspondingly more of the 2-year cycle is spent at the laboratory.

Design is accomplished using CAD (see below), but most of the early effort is put into a series of prototypes. These are uniformly named at Hitachi as follows:

are mainly used to fix ideas that have been previously worked out in the designers' heads, on scratch paper, and physically using prototypes. How the original concepts are generated and what role if any a computer plays are not easy to determine.

In answer to my question, they said that the main design challenges are weight, cost, part count reduction, meeting the specification, reducing the cost, and getting the production tooling and factory up to speed. These are, of course, just the concerns one would expect, but the order in which they were given might be indicative of descending order of difficulty.

final size and weight are difficult to discern. When the works designers begin converting the design to a manufacturable form, they apparently deal directly with the physical prototype itself rather than a computer rendition. A major tool available for this purpose is Hitachi's Producibility Evaluation Method (PEM) of which they are very proud. More on this below.

The Tokai works where these units are made is 2 hours away by train. This distance might be a barrier in a U.S. company but it is nothing to them. Meetings occur every week or two, with the majority at the laboratory during the first year, at the works during the second. This communication is “difficult" but apparently effective. Electronic communication is used mostly for sending text. Hitachi is aware of the advantages of being able to send engineering drawings between designers by computers, but right now such capability appears too costly in comparison to the benefits.

It is important to understand the scale of typical design projects when comparing use of computers and strategies for communicating between designers and works. In the car industry it is common for projects to involve

3,000 or more engineers. Here we are talking about a few dedicated people working feverishly 10 to 14 hours per day, with almost no assistance from draftsmen or technicians. These people will really own this design and will know every screw and hole.

Can the design cycle be reduced substantially below its current duration? In response to this they noted that use of the computer cut a short time out but in my opinion the amount quoted is not a lot. Could it be reduced by half? Their response is that the current duration is about the minimum, but this must be evaluated in the context of the kind of computer aids they are using and their outlook on what the future holds in CAD. As discussed below, this view is somewhat narrow. The computer saves them from timeconsuming mistakes such as interfering parts that must be made over. They claim that they will try to reduce the number of prototypes but they did not discuss or show me any concrete techniques for accomplishing this goal.

CAD Facilities

The mechanical design activities of I&MSL are supported by a CAD system comprising Hewlett-Packard workstations running HP's ME 30 solid modeling CAD software plus Structural Dynamics Research Corporation's (SDRC) I-DEAS solid modeler. Hitachi's contributions to this setup are a data translator for exchanging ME 30 and I-DEAS files plus a finite element package called CADAS. In addition, they have a nonlinear finite element package called ADINA, purchased from a small company founded by Professor Bathe at MIT. ME 30 is used mainly for its ability to represent many separate solid models of parts and assemblies of them, whereas I-DEAS is used because its solid models can be linked to analysis software such as FEM and plastic mold filling simulations. Apparently no single design package

The typical data flow of a design begins with a simple two-dimensional layout in Hitachi software called GMM which is sent via local area network to the HP machine for conversion to a rough solid model in ME 30. Some animation of the motions is done by using a separate computer to generate intermediate position data for moving parts and loading the data into ME 30 to create a series of views. No kinematic analysis software is used, although such has been available for almost 20 years. Hitachi will soon buy a package for this called ADAMS, which has recently been interfaced to I-DEAS.

After simulation, the rough solid model is refined into a complete model. Since there are 300 parts, this step can take a month or two. The model is broken into separate parts, on some of which FEM or mold fill analyses are done. A line drawing in 3D is then done and sent by LAN to another computer which makes conventional-looking drawings for the machine shop to use.

The above description is not intended to imply that 2D and 3D work are done sequentially; much work is done in parallel, with 2D being used for simple almost plane parts and 3D for complex parts or those with many surfaces.

In their opinion they could not design these VCRs without the 3D modeling. However, they admit that such tools permit only functional modeling and basic part fit studies. They cannot do tolerancing, kinematics, cost analyses, or their AEM. Reluctantly they admitted that they want to integrate AEM into CAD but they acknowledged in later discussions that they really do not have a technical approach for doing this.

Hitachi's View of the Product Development Process

My hosts from PERL made a presentation of the contents of a paper describing their approach to product

I&MSL was present at this time so I do not know the extent of their agreement with its contents or whether they follow this process. It is likely that the paper represents an idealization of a procedure that has emerged over a long period of time.

Simultaneous Cooperative Development

The process has both an organizational component and a technological one. The organizational component is called SCD (Simultaneous Cooperative Development) and is similar in spirit and methodology to Concurrent Design (CD), Simultaneous Engineering, and so on. However, SCD has developed “cumulatively,” whereas they see CD as a U.S. innovation implying a drastic new development that puts SCD on computers and makes heavy use of computer technology, design aids, and communications. Hitachi is presently surveying CD methods and research world-wide, especially in the United States. "Everyone in Japan sees the need to do SCD better and faster and it seems that the United States has more researchers working on this. Only a few Japanese companies see how to do it."

Hitachi admits that the idea of SCD is easier to define than to implement. It requires cooperation between marketing, design, and manufacturing to create a design that balances the goals of each group: utility, assurance of performance, and producibility. "Conflict breaks out as soon as the project starts." The spirit of cooperation is instilled by defining the ultimate user of the product as the one they are all working for. This is a good slogan but there is nothing to back it up in their methodology except to hold lots of meetings. The only clear guideline is that the project leader is someone assigned by the factory that will make the product, and this person, similar to the heavyweight defined by Clark and Fujimoto, draws on the skills