manufacturing knowledge and capabilities. Knowledge is defined very broadly as including expertise, test data, past designs and their field performance, deep engineering understanding, catalog data, government regulations, and company standard practices and design rules. The companies do not yet think in quite these terms and currently see wide area distribution of, and common access to, existing conventional CAD data as their main problem.

Structure of the
CAD Industry and
Technology Transfer Routes

Japan does not seem to have a CAD industry such as the United States or Europe has. Companies comparable to Computervision, Structural Dynamics Research Corp. (SDRC), and Mentor Graphics do not exist. No Japanese workstation has yet gained the popularity of those made by Sun or HewlettPackard. Japanese companies have taken three routes to obtaining CAD and CAE: buy hardware and software from the United States, invest over decades in writing their own software, and a hybrid of these. Companies that write their own face the problem of longterm support as technology and needs change, such as migrating to new hardware and integrating new programs into their existing software. But they have developed at least near-term solutions for these. (Nissan and Toyota have joint ventures with IBM and Unisys to support their home-grown CAD and sell it to their suppliers.) Companies that buy from the outside must accept their vendors' solutions to these problems and most have switched vendors at least once, a painful event. New switches will occur as workstations replace mainframe computers.

As research produces new CAD/ CAE/CAM tools and methods, an efficient technology transfer route needs to be developed. The rich companies

that write their own can support their own university research and insist on compatibility. Those that buy must rely on the software vendors to keep aware of research results and incorporate them into their product line. This route will not provide CAD buyers with the competitive advantage they need unless a variety of easy knowledge capture, data management, and software integration tools is also developed. Thus useful design research must be broad enough to recognize these as allied and essential issues.

U.S. and European companies face the same problems, of course, so solutions developed in Japan will be of great interest to everyone.

CURRENT PRESSURES ON JAPANESE INDUSTRY THAT AFFECT DESIGN PRACTICES OR SUGGEST

RESEARCH ISSUES

Japanese society is changing rapidly. Some trends visible now are recent while others date back years or decades. Those that follow were brought to my attention during my visit but are echoed in many publications, most recently in Reference 3.

Labor Shortage

Japanese industry has faced chronic shortages of factory floor labor since the early 1960s, when it was forecast the early 1960s, when it was forecast that the gross national product (GNP) would grow faster than the population. This has forced pervasive and relentless automation onto Japanese industry, which has, in turn, forced a revolution in how products are designed. Now the shortage has extended to engineers and scientists. Indeed, the birth rate is now 1.53 children per couple, and Japan faces falling population in a few decades. To maintain the standard of living, automation will have to extend to the field of design. Companies are

seriously worried about capturing the experience that they now boast of and converting it to computer form so that junior engineers can do the work of rapidly vanishing senior people.

Deterioration of

Lifetime Employment

Professionals are starting to discover the advantages of job mobility. At least 5% of employees voluntarily change jobs each year and the number is rising. Headhunter firms are springing up. An important result is that future design activities may not be carried out by people who have known each other for years. Risky design methods like task overlapping (see below) are likely to suffer. Computer design aids could help but the companies do not yet want to place strong reliance on them.

Shortening the Design Cycle

At the same time as the number of engineers is threatening to fall, the measure of competition has become rapid introduction of new products and new versions of existing ones. Companies have responded by intensively studying their organization and design process, automating key portions of the process, innovating management methods, and overworking their engineers. Some are having such trouble recruiting new hires that they give new graduates vacations to keep other companies from locating them. One company gives a car to each new hire. Several companies I visited spoke openly of lengthening the design cycle, possibly by silent industry-wide mutual agreement. Others resort to cosmetic redesigns, postponing more thorough efforts. While several companies presented their computer design aid activities as being targeted at shortening the design cycle, others de-emphasized the issue because it is politically sensitive.

Work Style, Work Life

As mentioned above, Japanese companies overwork their engineers, who in turn cannot decide if loyalty or resentment is the right reaction. Younger people are starting to rebel, and the government wants the working year reduced to the U.S. average of 1,800 hours by 1993, from the current average of 2,000 (2,200 or more in the high tech industries). A friend termed Japan a "herd society": once a trend starts, everyone joins in. If young people abandon engineering or refuse to work as hard as the previous generation, a crisis will occur.

Globalization

Japanese companies are finding that they cannot export their management methods to their branches in the United States and Europe. Overlapping of design activities is a risky approach since it requires starting a job before all the necessary information is available. Careful structuring of the design process, identification of the crucial information, and steady, deep communication between designers are required to keep serious errors from occurring. NonJapanese engineers are not used to such communication, they shun the risks of this approach, and they do not work long enough hours to accomplish it. Few companies outside Japan study and improve their design practices. Shared design projects, therefore, become uncoordinated as the foreign parts slip behind schedule. Research in design methodologies to overcome this kind of problem is not being attempted to my knowledge.

THE MAIN INTELLECTUAL ISSUES

How Japanese Companies Approach Product Design

Definition of a Manufacturing Company. Several Japanese companies Company. Several Japanese companies take a total view of their existence as manufacturing companies. They not only develop their own CAD software but also the most critical elements of their manufacturing and assembly equipment. On the other hand, they buy many of the components that go into their products. This keeps design staffs small (see below) and focuses the company on the essentials. That is, they are vertically integrated in the essentials of product realization and see this endto-end capability as a major competitive strength. U.S. companies are often vertically integrated in components and tend to buy their manufacturing and design facilities from a fragmented and undercapitalized vendor commuand undercapitalized vendor community. The Japanese approach reveals a stronger commitment to internal commitment to internal manufacturing excellence and provides vastly better opportunities for communication between product and process designers. It also provides funds for new process development and drives ongoing learning of better product realization methods and technologies. As one person put it, “You learn by trying, not by buying.'

Consistent with this commitment, many Japanese companies maintain production engineering as a corporate headquarters activity; it is often represented by an executive director, equivalent to an executive vice president in a U.S. company. Thus production engineering has a strong voice at the very top of the company. U.S. companies are often product-line oriented. Each product division has a voice at the top while production engineering is a function located at each factory. Its job is often merely to maintain purchased equipment.

Toyota and Nissan provide most of their own CAD software, while Nippondenso provides a significant portion of its. Sony, Hitachi, SeikoEpson, Fujitsu, and Nippondenso make their own robots (over 3,500 at Nippondenso and currently increasing at 1,000 per year). Matsushita makes its own circuit board assembly equipment. Some of this equipment enjoys strong outside sales, strengthened by in-house experience with its use. Sony has trouble selling its robots, and Nippondenso doesn't bother selling. Toyota and Nissan have commercialized several of their CAD programs, but only for the purpose of getting their suppliers to use them, not for general sale. Data and software compatibility is the goal.

Most of these activities date from the mid 1960s to early 1970s and appear to be unbroken, growing programs with long-term perspective and full top management initiative or support. Major objectives are set (extend ability to automate while attaining xx level of flexibility, or permit early detection of the most time-consuming kinds of design errors or uncertainties in car body engineering), and bit by bit they are attacked over many years.

Toyota, Nissan, and Nippondenso appear to have long-term strategies for allocating resources to computerization of the design process. For Toyota and Nissan, the focus is on the engineering-intensive and timeconsuming process of body styling and engineering, which normally suffers from huge data requirements and much trialand-error. For Nippondenso the focus is on supporting both routine mechanical design and breakthrough productprocess design for flexible production. Trial-and-error is not a big issue.

All three seem to favor achieving some level of end-to-end integration from concept to production engineering using admittedly approximate methods rather than delaying integration while perfection is reached in each of

the calculation steps in this process. All companies visited also recognize the need to provide all engineers with access to computers. The ratio of engineers to terminals varies from 5:1 to 3:1 with 2:1 or 2.5:1 being considered optimal.

Smaller companies naturally cannot afford such activities, but many in the range of 13,000 to 35,000 employees make their own CAD software and most in the 4,000+ employee range make key manufacturing equipment. An interesting exception is Mazda, which is selling off machine tool and transfer line divisions and using the funds to "in-source" some high tech, high valueadded components that they once bought. This is the route of "survival" in their view.

Systematic Approach. Every company I visited has a systematic, step-bystep plan for how products are designed. This is typical and not surprising. There is often a set of two to six prototypes spaced out at intervals during the process. Companies differ on when is the right time to introduce manufacturing and cost constraints and when to involve manufacturing engineers and factory personnel. The prototypes are often given names like “research," "function," "manufacturing," and "preproduction." When the product is a very new one, such as a videocamera, computers play a limited role (e.g., verifying the precision of tape threading) until the function prototype is finished, at which point CAD is used to document the design and support further engineering. When the product is an ongoing type, such as a car, the most advanced companies design the first prototypes directly into a CAD system. This statement applies to both totally new body styles and rather repetitive suspension components.

At Nissan, the first prototypes are built at the design center, whereas the last are made at the factory by manufacturing engineers or line workers. At

Hitachi, VCR mechanism designers build the first prototype with their own hands. At both companies, design responsibility shifts from the advanced design office to the factory's design staff beginning with the manufacturing prototype. At Nippondenso, the design process is so closely tied to increasing automation that process engineers are involved from the first day so that the necessary novel process methods can be developed. At Sony, product function designers are led by someone with at least 10 years' experience, and they take account of assembly sequence and assembly-related tolerances during functional design. Hitachi and Nippondenso have each evolved rather different design evaluation techniques for improving assembly. Neither has integrated them with CAD but both would like to. Companies disagree widely as to whether functional designers should be equipped with computer tools to critique manufacturability and assembleability of their designs, or whether these tools should be used by process engineers. Sony and Hitachi take the former view, while Toyota and Nissan take the latter. The difference perhaps reflects the different time scales for design (only a year or two for videocameras versus 4 years for cars).

The more sophisticated companies constantly review their design practices, including their deployment of computers. At Ishikawajima-Harima Heavy Industries (IHI), the process is being restructured using the critical path method (CPM) in order to cut the time. The idea is to carefully identify the information that each design step needs from prior ones and provides to later ones, plus when that information is needed or available. The information is ranked by importance or leverage and only the most important items are included in the CPM analysis. A tight flow of the most crucial information thus can be used to resequence the steps to produce a faster process. While no other company cited formal analytical

techniques, most are involved in ongoing or recently launched reevaluations of their design methods with the aim of reducing either cost or time or both.

These "restructurings” and “reformulations" of the design process are intellectually challenging and involve defining new work styles, data requirements, and software support requirements. "We used to buy software and adapt our work style to it," says a CAD director at Nissan. "Now, we will define our next generation work style and obtain or write software to suit."

Many American companies went through "painful" reformulations of their engineering design methods in the early 1980s and typically report that they are now satisfied with the results. Japanese companies are never satisfied.

Integration of Engineering and Business. Many of the companies visited have identified a theme for their business that is reflected in their efforts to deploy computers and other automation. At Nissan, this theme includes world-wide design activities with uniform standards, techniques, and supporting software. At Nippondenso, the theme is to conquer product diversity efficiently in a mass-production environment using a combination of product and process design. At Mazak it is to be the prime user of the manufacturing equipment it sells, both to gain experience and to act as a living laboratory for its customers. The machine tools it makes, interestingly, are rather ordinary, and their design is supported by only the most basic CAD. But Mazak's manufacturing automation is among the best and most well thought out in the world, and several aspects of product design are aimed at maintaining that excellence.

The consumer product companies recognize that marketing and product design are tightly linked. Nippondenso is especially good at identifying ways to design its products to meet the rapidly

varying product mix of its biggest customer, Toyota. Toshiba's laptop computer designers spend part of every week going over customer inputs so that new designs will be well received. Top executives set the specifications for the new product. At Nippondenso, the speed of the design process and the overlapped task method require top management involvement and fast decisions throughout the design process.

The long-term trend toward, and competitive advantage of, smaller, lighter, and quieter products (computers, cars, and everything between) is driving companies into more CAD and CAE. Strength, noise, and vibration characteristics of products are more critical. Lighter parts have thinner walls that vibrate or magnify noise more than heavier ones. Extensive finite element analyses are the only design tool available. Supercomputers and super workstations are being increasingly recruited.

A major theme running through this report is how to automate in the face of rapid changes in product technology and market shifts. Companies want to automate because automation is more consistent and efficient and produces higher quality than people can. But companies are afraid of being trapped with useless equipment if the product or the market changes. Researchers see this as either a problem of scheduling existing types of equipment or of improving the general technological level of equipment. Companies see it as a problem of product design, which often requires specific new production technology. The researchers' approaches are too narrow, but the companies' approaches, while more balanced and effective, lack generality. However, lack of generality may be inevitable and may never bring progress to a halt.

Integration of Product and Process Design. Japanese companies have known for years what U.S. companies once knew and apparently forgot, namely, that product and process design

need to be carefully coordinated. Until a few years ago, there was no special name for this in Japan; it simply was a fact expressed by the multidisciplinary composition of design teams. Now the names "simultaneous engineering," "concurrent design," and "concurrent engineering” have come into use. In the United States these are associated with attempts to apply computers to achieve this integration. The Japanese are puzzled by this development and wonder if it is something new. Their ability to assess U.S. activities is limited and I was questioned repeatedly on this point.

Many companies actively fear an invasion by computers into their human communication methods, thinking the United States will catch up and that computer communications will be too weak to support the intensity that Japanese currently achieve. (See below for discussion of small design teams.)

Success at product-process integration requires identifying just what information the downstream process designers will need from the upstream product designers, and vice versa. In the absence of a structure for this data exchange, integration degrades into arguments and confusion. While the priority is usually given to achieving the desired function, some Japanese companies are now so sophisticated that they can tailor product designs to favor some very efficient and flexible manufacturing methods without impacting performance at all. Nippondenso is the best of these among companies I the best of these among companies I visited. The research community has barely recognized this issue. Potential approaches include information analysis of design processes, cost structure analyses of fabrication and assembly, and modularization methods for products.

Overlapping Tasks. Overlapping design tasks, described above, presents numerous problems. Some academic researchers in engineering predict that overlapping can be used only on

repetitive products like cars where there is a well-developed design process in place. However, Nippondenso claimed that it uses this method when developing quantum step improvements in existing products, an effort that means total redesign and many new manufacturing and assembly processes.

To support overlapping with analytical and computer methods requires creating ways to systematically detect and structure data and information flows in ways that are more sophisticated than IHI's methods for resequencing. Right now, all the companies depend on communication between engineers who have worked together for years. They can anticipate each other's actions and compensate ahead. For example, a stamping die designer can peek at a car panel design and see trouble in one region. He can then leave this region blank in his die design or can have the die made with extra metal in that region that can be removed later when the precise body shape becomes known. Often stylists will not permit outsiders to peek at their unfinished designs since they do not want to be blamed if the design changes later. Long association creates the necessary trust.

One should not conclude that the companies depend solely on this unstructured communication. In fact, Toyota insists that all information release is approved, but incomplete or preliminary information can be released with only low level approval whereas final information requires high level approval. Mazda has a highly structured set of over 20 design reviews that guarantee input from and information for all the relevant departments.

Small Design Teams. At many companies, the number of designers and engineers assigned to one product seemed small. For example, the rough statistics in Table 1 apply to product function designers designing new products (not minor redesigns of existing ones) at several companies visited.