« iepriekšējāTurpināt »
Therefore, it is now accepted that versus 1,000/month). Thus the differ- topic and may indeed represent a new decisions made during design domi- ence must lie in the extent of the changes: way of thinking (Ref 5). On the other nate the cost and performance of the apparently they can be rather large and hand, Japanese designers, when forced product, and most downstream deci- disruptive in the United States whereas to do things faster, may have thought sions are subordinate with little ability in Japan communication has ironed the problem through and made priorto affect cost or performance. Thus out the big problems early and only ity lists of what information they really intense attention is being given to how small ones crop up.
need early in the die design process. to accomplish CD. Even Toyota and The difference between regional They may have also found out that such Honda, according to Fujimoto, feel the performance here lies in the organiza- information may be easier to obtain need to do this better.
tion of the design process, the timing of than they thought, especially if they Honda used to have a separate design information transfer between phases, promise not to complain if the design "company” called Honda R&D, which and the type of information transferred. changes later. In fact, Fujimoto claims essentially sold finished car designs to Japanese production tooling designers that overlapping of the process steps Honda Motors. The manufacturing can use sketches, line drawings, or came first, as a means to shorten the system engineers at Honda Motors preliminary drawings to begin their own cycle, and communication lines develprided themselves on their flexibility, designs. In the United States, tooling oped in response. (See the Toyota their ability to accept a lot of "wild designers are reluctant to use such reports, where the claim is made that changes at the last minute," though not information and body designers are information release is tightly controlled after the product was launched. My reluctant to provide it because changes and does not occur freely between longown opinion is that in the United States give rise to blame-fixing and finger- time acquaintances.) the wish of manufacturing engineers to pointing. Furthermore, some informa- As another example, in the aircraft appear flexible has been a barrier to tion is not directly obtained in Japan engine business, it has been found that increased CD. Until the early 1980s in but rather is intuited in the sense that the most useful advance information the United States, the quickest way to designers know each other and their for tooling engineers about an engine be swept aside was to complain that the styles, as well as the most change-prone shaft is its length and outer diameter. design needed to be changed. It is pos- regions of a design. Thus they can antic- This information permits them to order sible that Honda R&D's designers knew ipate some changes by providing safety raw material and begin machine design. enough about producibility that they factors, such as excess metal on the die However, the shaft's inner diameter could avoid a lot of obvious problems. face that will be removed later when does not strongly influence prelimiU.S. designers are less likely to have final designs are in hand.
nary machine design and can be learned such knowledge.
Strangely, much of the information later. Also, preliminary finite element
of most forward use to die designers method (FEM) analysis can be done Particular Fujimoto/Clark exists when the final clay models are knowing only a rough inner diameter, Findings
finished, which is before any body panel and this analysis helps determine what
design begins. According to Fujimoto, the final inner diameter must be. In automobile companies, the time the die designers “lack access" to the Another important factor is a to design and field a car model is dom- clays, which can be many miles away, or company's reward structure and maninated by the time to design the body locked behind elaborate security bar- agement culture. Companies that penaland the stamping dies needed to make riers. Another source of such informa- ize design changes may stifle early the body parts. This time is, in turn, tion that is not used is the model shops communication and overlapping of dominated by the time to design, make, where preliminary dies are designed. processes and force designers to furand test the dies themselves. In Japanese Habit plays a big role in such lack of nish only “final” designs. companies, the lead time for dies is information transfer. about 1 year versus 2 years in the United Die designers are not used to being Human Communication States and Europe. Furthermore, the asked what information they would need Versus Computer impact of a die design change on both and thus have not taken the time to Implementations cost and schedule is much lower (10% decide what would be of the most use. to 20% excess cost over budget versus They take it all in at once without par- Fujimoto/Clark represent a research 30% to 50% in the United States and titioning it in their minds or in their paradigm that is not surprising in a Europe). On the other hand, the abso- design process. Formal methods to business school, namely, that managelute number of engineering changes is partition a design process are only ment techniques and human activities not too different, he says (500/month recently drawing attention as a research dominate, and computer tools and other
advanced design and manufacturing that will permit all the data describing problems, integrating engineering and technology are subordinate. Toyota a product to be converted and commu- business data, and so on. Much of this seems to be living proof of this. Fujimoto nicated within and between factories, kind of information is currently extends this claim by pointing out that design studios, suppliers, vendors, and embodied in the experience of each many aspects of car design simply can
company's engineers, meaning that new not be handled long distance by any Fujimoto argues that communica- kinds of software tools and knowledge existing technology, such as electronic tion capability is an organizational skill capture methods will be needed. Commail or real time links between com- that takes years to build up. It is based panies that put effort into this kind of puter workstations. “Only face to face on personal contacts developed over computing will definitely gain a comcommunication will suffice.” Examples years of working on similar design petitive advantage that others will not concern how to capture certain human projects and is fortified by the stability be easily able to copy. expressions in the design, such as the of employment and length of job assign- The willingness of companies to feel of the suspension or the handling, ments typical of Japanese companies. invest internally in computerized design or even the sound the engine and exhaust By contrast, he views computers as infrastructures, including supporting system will make (so-called esthetics of commodities that anyone can buy and research on advanced design tools and engines mentioned above). Fujimoto install quickly. Thus no competitive methodologies, may be an essential Clark posit the existence of “heavy- advantage is gained by going this route, factor in future productivity and comweight product design managers” as although anyone who does not will be
although anyone who does not will be mercial health. This will be especially the main carriers of this type of infor- left behind. But adopting the right important as cadres of experienced mation (Ref 6). Such managers are
communication methods is not a com- engineers retire and take their experideeply experienced in both market and modity and others cannot quickly jump ence with them. engineering aspects of car design and in and achieve the same results. have been raised from a background in The engineer paradigm people view Final Comment engineering. Their use has been adopted the situation exactly oppositely. There in the car industry world-wide as car is increasing evidence in U.S. companies Clark and Fujimoto have focussed designs have become more sophisti- that communication and overlapping on one industry, automobiles, one aspect cated and complex. Heavyweights do of product design and production equip- of car design, some moderately comnot simply push schedule and budget ment design can be introduced quickly plex parts with aesthetic elements, and but meet with designers every day, and effectively. Methods used so far one period of industrial history, the personally drive the cars on the test require intense personal pressure and 1980s. Other industries are charactertrack, talk intensely with the test engi- long working hours (same as in Japan!) ized differently, as are other items they neers, and so on.
and considerable fatigue results. This make. Electronic component design is However, an alternate paradigm means that such methods have not been heavily computerized and data driven, exists in the engineering schools and in institutionalized yet, but they will be, at whereas almost no mechanical design engineering research on the design least in the survivor companies. is similarly structured. In another report, process. This paradigm states that However, computers may not be I describe a Japanese company's methadvanced computer tools are essential commodities after all. It is true that ods and equipment for designing small to the design of complex products, the workstations are becoming commodi- VCR mechanisms as well as their outprime examples being microprocessors, ties, and so is basic commercial design look for future competitive advantage. aircraft, and cars. The unification of software, such as CAD tools. However, While cars are designed by teams design data, production orders, market the true differences between companies numbering in the thousands, these VCRS information, and control commands to will soon shift: right now the difference are designed by a team of 10 engineers production machines is a major goal of is in how much CAD has been adopted. total. Communication problems clearly modern integrated manufacturing and In the future, the difference will be how vary with the scale of the project. Yet, is the subject of the emerging MS project much beyond CAD is achieved. This with 300 parts, VCR mechanisms are (Intelligent Manufacturing System) includes capturing design knowledge not simple. among others. Another similar effort from past designs in computer-readable Fujimoto added in later a commuis the international standards project form, making product design data avail- nication with me that the characteriscalled PDES (Product Data Exchange able to tooling designers, creating tics of the auto industry in the 1980s Standard using STEP). (STEP is the computer “design critics” that can auto- may not apply to the 1990s. Indeed, European acronym for PDES.) PDES matically comment on a product design computers and their constituent softwill establish a data exchange standard and alert designers to producibility ware and databases may become the dominant feature of leading companies, Hitachi's organization is a hybrid of Focus of the Visit especially those that have already product-line orientation and function improved their design productivity, orientation. Each business group (such The focus of this visit was Hitachi's learned how to overlap design tasks, as Consumer Products, Industrial Prod- approach to the design of complex and identified the knowledge and algo- ucts, Power Plants, etc.) reports directly mechanicalitems. Takahashi chose the rithms that must be captured in software. to the president. But there is a separate VCR, and it is a good choice. The tape
Finally, I might add that this is the Production Engineering Department changer-player mechanism typically has first report written while I was in Japan. that also reports directly. Finally, the 300 metal and plastic parts, stamped or My subsequent findings amply support nine research laboratories also report injection molded. Most parts are riveted the main points raised in this report: directly. This means that the president together, although there are a few very the advanced companies indeed are directly can control research, market- small screws. This is a delicate, precise building on their accumulated experi- ing, and production engineering and mechanism which must be rugged and ence, capturing their efficient manage- can in principle mediate disputes among reliable while at the same time handling ment methods and experience in com- them. These are among the deepest delicate tape, being lightweight, easy to puter tools. See especially the reports disputes that can occur in a diverse assemble, and low cost. Similar prodon Toyota, Nissan, Mazda, and technically sophisticated manufactur- ucts are made by Sony, Matsushita, and Nippondenso. ing company.
In many Japanese companies, there Products with similar characterisCAD AND PRODUCT
is an executive vice president of pro- tics include compact disk players, DESIGN METHODOLOGY duction engineering, attesting to the cameras, and miniature hard disk drives AT HITACHI
importance these companies give to for laptop computers. Such products
manufacturing excellence. In most U.S. are often called "mechatronic" to call 11 June 1991
companies the organization is strictly attention to their hybrid mechanical
by product line, with each factory hav- electrical-optical-computational charBackground
ing its own manufacturing engineers. acter. Major issues in their design include
These people usually have little say in deciding whether to embody a function I visited Hitachi Yokohama works the design of the product and often optically, electronically, computaon 11 June. My host was Mr. merely take care of equipment that the tionally, or mechanically. Their fabriMichio Takahashi of the Production company has purchased outside. Dif- cation and assembly involve cleanliEngineering Research Laboratory ferences like these contribute to rela- ness, high tolerances, as few adjust(PERL). The visit focussed on design tive strengths and weaknesses of man- ments as possible, and extreme efforts of VCR mechanisms and camcorders ufacturing companies, in my opinion. at uniformity and quality of output. An and our hosts were designers in the At I&MSL there are 350 researchers, additional complication is that all such Image and Media System Laboratory of whom 70% have an electronics back- products are extremely small and get(I&MSL), which is next door to PERL. ground while the rest are mechanical, ting smaller.
The I&MSL is a product develop- physical, and chemical. Products include My goal in visiting was to see how ment laboratory within the Consumer VCRs, camcorders, optical disk stor- Hitachi meets these challenges, what Products Division. Its role is to create
age systems, high definition TVs computer aids if any are being used, new technology for products as well as (HDTV), digital audio tape systems, and what future computer aids Hitachi to design new kinds of products embody and their associated electronics. The thinks it needs. Additional goals were ing such technology. Products are laboratory has extensive large scale to see where Hitachi obtained the manufactured at factories (called works) integration (LSI) design and test fab- computer tools it has and to see whether and such works also have design depart- rication facilities and obviously puts it plans to develop its own in the future, ments. The latter make small updates most of its efforts into the electronic deal with universities, or buy from to existing designs and lead in the con- aspects of design.
commercial companies. version of preliminary designs into Most of the above information and manufacturable ones. PERL is one of what is reported below are illustrated Summary of Discussions nine research laboratories, others deal- (though sketchily) in a brochure they and Tours ing with energy, basic research, mechan- gave me about I&MSL. ical engineering, systems development,
I met in the morning with Dr. microelectronics, design, and “advanced
Yoshihiko Noro, Chief Researcher, and research."
several of his main assistants. We
discussed the product design cycle for A totally new product may absorb are mainly used to fix ideas that have miniature VCRs, including how long it 2 years of advanced development of its been previously worked out in the takes, what the chief obstacles are, and basic technologies before any specific designers' heads, on scratch paper, and who does what. I was also shown the product design begins. Then it takes
product design begins. Then it takes physically using prototypes. How the CAD center.
about 2 more years to create a product. original concepts are generated and After lunch, I got a brief tour of For a relatively mature product, one of what role if any a computer plays are PERL's robotics laboratory where I these years will be devoted to design not easy to determine. saw several things I had seen in earlier development at the laboratory while In answer to my question, they said stages of development on previous visits the other year will be spent converting that the main design challenges are (six in all since 1974).
it into a manufacturable item utilizing weight, cost, part count reduction, Finally, there was a presentation by the works designers. For less mature meeting the specification, reducing the PERL of a recent paper on their products, correspondingly more of the cost, and getting the production toolapproach to Concurrent Design (see 2-year cycle is spent at the laboratory. ing and factory up to speed. These are, the report on my visit to Prof. Fujimoto) Design is accomplished using CAD of course, just the concerns one would and an open discussion on this topic. (see below), but most of the early effort
(see below), but most of the early effort expect, but the order in which they We planned several additional visits, is put into a series of prototypes. These were given might be indicative of including to factories where washing are uniformly named at Hitachi as descending order of difficulty. machines and automotive components follows:
The early prototypes are naturally are made. These visits are covered in
very clunky and heavy, with many hand• First prototype--preliminary design. made parts. Function can be tested but
final size and weight are difficult to VCR Design Cycle
• Second prototype--the main focus discern. When the works designers begin
is to achieve a design that fits the converting the design to a manufacturI&MSL is responsible for consumer size required by the market goals; able form, they apparently deal directly stand-alone VCRs, VCRs inside cam- the I&MSL engineers use Hitachi's with the physical prototype itself corders, and professional studio VCRs. assembly evaluation method dur- rather than a computer rendition. A The latter are their newest products ing this phase.
major tool available for this purpose is and are made at the glacial rate of 10
Hitachi's Producibility Evaluation per month. By contrast, the others are • Third prototype--responsibility of Method (PEM) of which they are very made in the hundreds of thousands per works designers. The main focus is proud. More on this below. month. Naturally, the design of some- cost reduction; in this prototype,
The Tokai works where these units thing made at 10 per month is quite Hitachi's assembleability evaluation are made is 2 hours away by train. This different from 100,000 per month, since method (AEM) is used in coopera- distance might be a barrier in a U.S. automation of the latter is almost a tion with the production technol- company but it is nothing to them. necessity. Thus each part must be very ogy group.
Meetings occur every week or two, with simple, must be able to be picked up by
the majority at the laboratory during a simple gripper, and installed in a • Pre-mass production prototype --30 the first year, at the works during the simple straight-line motion.
units made on production tooling second. This communication is “diffiThese VCRs are an incredible jum- and tested for performance, then cult” but apparently effective. Elecble of stacked, nested, and intertwined 100 more tested for quality assur- tronic communication is used mostly levers, rollers, springs, sliders, cams, ance and final checkout of the tooling for sending text. Hitachi is aware of the and bars, run by a motor and several
advantages of being able to send engilittle rubber belts. Most of the parts I was shown several such prototypes. neering drawings between designers by occupy the bottom 3-mm-thick layer of Computers are used to lay out the computers, but right now such capabila unit that is about 2 cm thick, 10 by mechanical parts, check for interfer- ity appears too costly in comparison to 14 cm. Design is therefore an exacting ence during functional motions, see if the benefits. process requiring fitting parts into straight line assembly motions are It is important to understand the small spaces as well as determining possible, perform stress analyses, and scale of typical design projects when how the moving parts will travel while avoid “simple” and “obvious" errors comparing use of computers and stracarrying the tape from the cassette to that would seriously delay the design. tegies for communicating between the read head.
See below for details of computer facile designers and works. In the car indusities. It appears that computer models try it is common for projects to involve 3,000 or more engineers. Here we are can do all the things Hitachi wanted, a development (Ref 7). No one from talking about a few dedicated people common situation.
I&MSL was present at this time so I do working feverishly 10 to 14 hours per The typical data flow of a design not know the extent of their agreement day, with almost no assistance from begins with a simple two-dimensional with its contents or whether they foldraftsmen or technicians. These peo- layout in Hitachi software called GMM low this process. It is likely that the ple will really own this design and will which is sent via local area network to paper represents an idealization of a know every screw and hole.
the HP machine for conversion to a procedure that has emerged over a long Can the design cycle be reduced rough solid model in ME 30. Some period of time. substantially below its current dura- animation of the motions is done by tion? In response to this they noted using a separate computer to generate Simultaneous Cooperative that use of the computer cut a short intermediate position data for moving Development time out but in my opinion the amount parts and loading the data into ME 30 quoted is not a lot. Could it be reduced to create a series of views. No kine- The process has both an organizaby half? Their response is that the cur- matic analysis software is used, although tional component and a technological rent duration is about the minimum, such has been available for almost one. The organizational component is but this must be evaluated in the con- 20 years. Hitachi will soon buy a package called SCD (Simultaneous Cooperatext of the kind of computer aids they for this called ADAMS, which has tive Development) and is similar in are usingand their outlook on what the recently been interfaced to I-DEAS. spirit and methodology to Concurrent future holds in CAD. As discussed below, After simulation, the rough solid Design (CD), Simultaneous Engineerthis view is somewhat narrow. The model is refined into a complete model. ing, and so on. However, SCD has computer saves them from time- Since there are 300 parts, this step can developed “cumulatively," whereas they consuming mistakes such as interfer- take a month or two. The model is see CD as a U.S. innovation implying a ing parts that must be made over. They broken into separate parts, on some of drastic new development that puts SCD claim that they will try to reduce the which FEM or mold fill analyses are on computers and makes heavy use of number of prototypes but they did not done. A line drawing in 3D is then done computer technology, design aids, and discuss or show me any concrete tech- and sent by LAN to another computer communications. Hitachi is presently niques for accomplishing this goal. which makes conventional-looking surveying CD methods and research
drawings for the machine shop to use. world-wide, especially in the United CAD Facilities
The above description is not intended States. “Everyone in Japan sees the
to imply that 2D and 3D work are done need to do SCD better and faster and it The mechanical design activities of sequentially; much work is done in seems that the United States has more I&MSL are supported by a CAD sys- parallel, with 2D being used for simple researchers working on this. Only a few tem comprising Hewlett-Packard work- almost plane parts and 3D for complex Japanese companies see how to do it.” stations running HP's ME 30 solid parts or those with many surfaces. Hitachi admits that the idea of SCD modeling CAD software plus Struc- In their opinion they could not design is easier to define than to implement. It tural Dynamics Research Corporation's these VCRs without the 3D modeling requires cooperation between market(SDRC) I-DEAS solid modeler. However, they admit that such tools
However, they admit that such tools ing, design, and manufacturing to create Hitachi's contributions to this setup permit only functional modeling and a design that balances the goals of each are a data translator for exchanging basic part fit studies. They cannot do group: utility, assurance of performance, ME 30 and I-DEAS files plus a finite tolerancing, kinematics, cost analyses, and producibility. “Conflict breaks out element package called CADAS. In or their AEM. Reluctantly they admitted as soon as the project starts.” Thespirit addition, they have a nonlinear finite that they want to integrate AEM into of cooperation is instilled by defining element package called ADINA, pur- CAD but they acknowledged in later the ultimate user of the product as the chased from a small company founded discussions that they really do not have one they are all working for. This is a by Professor Bathe at MIT. ME 30 is a technical approach for doing this. good slogan but there is nothing to used mainly for its ability to represent
back it up in their methodology except many separate solid models of parts Hitachi's View of the Product to hold lots of meetings. The only clear and assemblies of them, whereas I-DEAS Development Process
guideline is that the project leader is is used because its solid models can be
someone assigned by the factory that linked to analysis software such as FEM My hosts from PERL made a pre- will make the product, and this person, and plastic mold filling simulations. sentation of the contents of a paper similar to the heavyweight defined by Apparently no single design package describing their approach to product Clark and Fujimoto, draws on the skills