« iepriekšējāTurpināt »
the computer's suggestions being real Research Laboratory: Americans jump strategy. Thus it was interesting to hear is enhanced if they are built up
from too quickly into system software (equiv- this topic discussed by him and two verified elements.
alently are too top-down oriented) researchers from Hitachi, Mr. Ohashi Thus his model is that the designer without thinking the process through and Dr. Taniguchi. Both have advanced can manipulate such elements in what first, whereas Japanese are too slow degrees from U.S. universities and are he calls a "virtual factory" that con- and spend too much time figuring out active in PERL developing software to tains models of machining, molding, their processes and perfecting them aid manufacturing engineering, assembly and assembly processes. These models manually before daring to put in any evaluation, and concurrent engineering. may be approximate but that would be software (too bottom-up oriented). The Hitachi people reflect the good enough for preliminary design. Iam also reminded of the difference opinion of the Nissan people, namely,
in approach between Toyota and GM that “90% of problems in product design Aside on Model Accuracy regarding die design software. Toyota methods can be solved by management,
has a series of elementary analyses that and only 10% can be solved by comThe idea that a rough model with the designer can ask the computer to puter tools and computerized knowlimprecise or incomplete data should performon a dieshape. These are quite edge bases.” They base this opinion on be sought for early design is consistent approximate but can avoid every known observing their company designing really with Toyota and Nissan being willing disaster. The totality of the programs is new products for which there are no to launch downstream design tasks when an end-to-end system for turning designs established procedures or tools. Espeonly partial information is available. into finished dies. GM critiques this cially important is the lack of correBody styling data accurate within 1 mm system (privately to me and to others) sponding manufacturing processes. are available months before the last as not having any real engineering Experiments, communication, redesign, mm is pinned down. Yet nearly all of analyses or accurate simulations of the and feedback are the essential elements, the design of dies can be done within metal deformation process. GM has and the success of these is mostly influ1 mm and finalized at the end. None of delayed implementing a similar system enced by management. the stamping simulations is likely to be pending completion of accurate models, The main problem in complex design affected by a 1-mm change.
a step Toyota gave up on immediately is that different groups or tasks have Thus a major component of research as requiring too much computer power. conflicting goals and there is no stanin support of early design should focus Thus Toyota can be said to value inte- dard way to work these out. The team on determining just what information gration over accuracy, while GM favors design method brings the issues out is of real value in launching any given accuracy over integration. Toyota sooner but that only causes embarrassstep in the design and the level of accu- appears to be ahead, although Kimura ment. Hitachi's big point of pride is the racy at which this information must be jokes that CAD-designed cars seem to “user first” slogan, which supposedly provided. Such a study will likely look alike.
focusses the team on the customer and reveal that much of the demanded
makes them forget their internal conperfection of data is not needed right Design and Manufacturing flicts. What solution to conflict X will away and that a preoccupation with Problems. In this area, there is little benefit the user the most? The fact that perfection and finality is a time waster work so far. It covers concurrent engi- design tools could explore many possiin many design processes. It also may neering, parametric design of routine ble answers faster is not appreciated. be a time waster in many searches for objects, machining simulation, design The reason may be that it does not good computer models of products and for machining, assembly process plan- seem to be a real option. It is just a processes.
ning, and standardization of CIM data. promise by researchers. He cited Toyota's method of evolv- The group is active in STEP and other In this context, what is concurrent ing design tools: make a simple tool standards activities.
engineering? Is it any different from that helps a specific design step; observe
simultaneous engineering? Are they how the designers use it and get their The Relative Importance of just the team design method with suggestions for improving and broaden- CAD/CAM/CAE and
computer support? Isn't concurrent ing it to cover more of the steps. This Management in Product engineering just something that way the tools grow organically and no Development Strategies Japanese companies have done for a useless tool is pursued or imposed on
long time? the designers. It is an example of a As a researcher in CAD/CAM/CAE, Discussion on these points reveals larger difference identified to me last Kimura feels that these topics are central little consensus except that all is not year by Mr. Hazeki of IBM Tokyo to any successful product development rosy inside Japanese companies. Some people refer to “big business syndrome” question:” How can the IMS project Background of Nissan and in which top managers want to inter- achieve its main goal, which is for Japan the Nissan Technical Center fere and decisions take too long to to relieve many international trade (NTC) make. Ohashi cites a recent Business problems by giving its advanced manuWeek article on concurrent engineer- facturing technology to the United States Nissan is Japan's second largest car ing as a clever way to convince man- and Europe? Why are there so many company, behind Toyota, and fourth in agers that communication is really political problems over IMS? Surround- the world. In the last several years important.
ing this question are his wonderment there has been considerable progress Kimura feels that computers can- that the United States can develop upgrading CAE and an effort is underway not by themselves shorten the design advanced ideas like CAD and concur- to link styling, design, engineering, and cycle although they can improve qual- rent engineering with computers and test data into one system world wide. ity and help improve the technological still not spread them around adequately. Kobayashi acknowledges that this will level of products. But shortening the In one sense Kimura answered the be difficult for two main reasons: softcycle is almost exclusively a manage- question himself but I also offered an ware incompatibility and people incomment factor. Thus concurrent engineer- answer. The United States is behind patibility. So far they have linked design ing can help shorten the cycle. Japan in its sophistication regarding and engineering offices in Japan, the
The Hitachi people feel that con- manufacturing. It is not clear if we could United States, the United Kingdom, current engineering could help improve adequately absorb Japanese technol- and Spain and can share data and softquality because they see the current ogy if it were “given” to us. Thus I said ware because (luckily, says Kobayashi) push to shorten the cycle as threaten- that successful international projects all these offices had IBM or IBMing quality. There is too little time to do must have well-defined technology compatible mainframes already when the necessary analyses and tests. Thus transfer mechanisms built into them. the integration effort began. providing knowledge bases, integrat- Such mechanisms must include long- The Nissan Technical Center is in a ing design and test data, and improving term hands-on contact with the tech- beautiful location and comprises many information flow are all ways that con- nology and cannot be confined to deliver- modern buildings. “Amenities are current engineering could help the ing reports or presenting papers. Such becoming more and more important." design process move faster.
conditions should apply to any kind of The main design/engineering building, Now that the United States seems "manufacturing” technology, includ- where we visited, is new since April to be catching on to the big secret of ing management techniques, computer and has an open, airy, atrium design. communication and team design, will software, robots, sensors, and so on. There are no offices, only wide open Japan have to invent something bet
spaces with many large tables where ter? They secretly worry about this but NISSAN TECHNICAL
meetings can be held. All the engineerhave no answer. It is the big debate CENTER (NTC)
ing disciplines for car design and proinside Hitachi right now.
totype production are housed in this 8 July 1991
building except the engine people, who Closing Comments
will move in next year. Thus, colocaBackground
tion has been adopted as a design effiKimura's view of design research is
ciency strategy both impressively broad and very prag- Our hosts were Mr. Jun-ichi All of Nissan's car design in Japan is matic. His topics seem to be aimed Kobayashi (Product Development centered at NTC or at the Central directly at the main needs of mechani- Section, planning Nissan's future CAD Laboratory in Yokosuka, where funcal design: higher levels of modeling capabilities), Mr. Yoshida (in CAE user damental studies are carried out in that include symbolic components such support), Mr. Katoh (Yoshida's boss), materials, power trains, electronics, and as constraints, maintaining the link and Mr. Ono (Production Engineering vehicle technology (body, chassis, user between design and manufacturing CAE). Kobayashi was the most infor- interface). In the styling and design processes, and striving to support design mative and interesting of these. He has area (perhaps half of all vehicle design in fuzzy situations like early design where been with Nissan since 1974 and spent and engineering?) the engineers are answers are needed but data are limited. 10 years in Manufacturing Engineer- working on several vehicle design proj
The IMS idea was born in this envi- ing before coming to NTC to work on ects at any one time. I estimate that 100 ronment and thus it has a very sophis- CAD/CAM/CAE.
to 200 stylists and designers are at work ticated intellectual base. Kimura sur
on one car project (plus probably another prised me by asking me a "very general
200 doing more detailed engineering). Unless I misunderstand badly, this is a The main reason why Japanese car terminals. But he admits that CAD/ remarkably small number of engineers companies can design faster is the “fuzzy CAM/CAE may have kept the lead time for the number of cars they design. design method." This means being from becoming even longer and the Uncounted in this total are manufac- willing able to start a design task before number of engineers from growing. I turing engineers at NTC and at the all the “required” information is in fact note that cars are very much more factories.
in hand. There is less structure in this complex now than in 1980.]
method and, yes, changes must be made It is worth noting, when considering Product Development Process later. But no one gets blamed for these (see below) that Nissan has distributed
changes and in any case they come amid its design software worldwide to its I estimate that the typical new car a stream of other changes anyway, so overseas design offices, that exporting development process takes about the important thing is to be able to the fuzzy design method has been a 4 years. Styling and design occur dur- respond quickly to changes. The CAD failure so far. Real problems thus remain ing the first 1.5 years or so. “The rest is system contains no explicit change trial, error, and design changes.” In the management system, according to Product Development Methods first prototype, new components are Kobayashi.
and Use of Computers often tested by attaching them to exist- Note that this explanation agrees to ing cars. Later prototypes use com- the letter with that of Prof. Fujimoto Nissan's CAD/CAM/CAE consists pletely new equipment. Die design in (see the Fujimoto visit report.) Even of many computers, programs, and particular takes about 1 year, but changes though Kobayashi is responsible for databases. Nissan is in the midst of keep coming in during the last year planning Nissan's future CAD needs, trying to merge all these into one seambefore production starts. During the he refused to ascribe much importance less process but it is a hard job. What first 1.5 years or so, or whenever proto- to the use of CAD/CAM/CAE in speed- they have is like “islands of automatypes are being built at NTC, the fac- ing up the design process. In view of tion” in the factory world. tory people make frequent visits to NTC. later examples (crash test simulations, Nissan is in the process of convertAs prototype work shifts to the fac- search for assembly problems before ing from styling using clay models to tories, the NTC people travel there. the first prototype, etc.), this claim is direct styling on the computer. In the CAD data do not travel well to the disingenuous. But I did not try too hard old clay model method, stylists carve factories (see below). to fight it.
quarter and full scale models and paint Design begins while styling is still in (Via a separate communication, I them to look quite real. These models progress. This is typical of overlapped obtained the following opinions from are digitized and computer data created. jobs in Nissan and other Japanese car Prof. Hiroshi Sakurai of Colorado State In the new method, stylists draw shaded companies. At some point during styling, University, a graduate of Tokyo Uni- sketches which “technicians” (high the exterior shape is frozen and this versity and former employee of Nissan. school graduates) put into the styling master data package is then passed on The main reasons why Nissan can design computer to create boundary patch to downstream activities. Interior styling a car in 4 years are: (1) average 2,700 models whose shape resembles that continues to be worked on after the working hours per year for engineers indicated by the shading in the sketches. exterior is frozen.
versus 1,800 per year in the United These models can be shaded and renApparently all Japanese companies States (Toyota people work even longer); dered very realistically, including showfeel pressure to reduce the design cycle (2) overlapping of jobs, for example, ing light reflecting off the surface. The but there is no specific program at Nissan Nissan's body engineering begins when stylists sit with the technicians and adjust to do so. “We already design cars in only rough exterior data are available the shapes of the surfaces. Apparently 4 years, compared to 6 in the U.S. and whereas U.S. designers wait for the technicians make a very strong 8 in Europe. If we shortened any more, smoothed data even though the differ
smoothed data even though the differ- contribution to this process and are we would just get a lot of negative ence is often less than 1 mm (Fujimoto not mere data input drones. They must reactions from overseas. So we will use cited precisely the same example!); have both artistic and computer skills the ability to design faster in other (3) outsourcing of many components as well as the ability to make the stylists ways." I was not told what ways, but to small companies who work their confident and comfortable in the presobviously two are to reduce the time engineers even harder and pay them ence of the computer. At some point in pressure on their engineers (who often even less. He notes that even in 1980 this process, the model is used to drive stay until 9 or 10 p.m.) and to do more Nissan could take a car from exterior a large gantry NC machine which carves analyses on each design.
freeze to volume production in 2 years out a clay model which is then painted even though they had only a few CAD and judged as in the old method.
When the stylists are satisfied, the For example, a torque wrench must In any event, if Nissan continues data become the master data for the have 60° swing room. Less requires along this track, it has the CAD tools in car's exterior and are never changed redesign or an OK from the factory. hand to do the necessary studies in thereafter.
Nissan tried using Hitachi's assemble- advance of building prototypes, an Interior styling is apparently done ability evaluation method (AEM) but essential ingredient. However, the long in a similar way.
concluded that it is intended for small term effort may not be directed at effectThe boundary model data are passed, parts assembly of large production rate ing robot assembly. Data in a handout perhaps with some pain, to the design items for consumers. Neither Nissan show that Nissan ramped up from 540 and engineering computer system. This nor Hitachi can do real design- robots in 1980 to 2,000 in 1985 but has system is separate and builds a new performance-manufacturability trade- only 2,462 in 1990, indicating real satdatabase starting from the master offs, they say. Cost should be the cri- uration at this level. “In 1980 we thought exterior data. Most of the work is done terion but no one has a method. robots would replace people. We know in 3D wireframe form using Nissan's (Ten years ago, engine compartment now that it won't be that easy. Yet the proprietary CADII system (see history design problems were found by making goal is still 100% automatic final below). This system is good for the Xerox copies of drawings of parts and assembly.” designers, who can put in as much detail sliding them around on a table, or by Stamping dies and plastic injection as they want without swamping the making transparencies of parts and doing molds for body and interior parts can data storage and manipulation the same.)
be designed in about a year. All the capabilities of the software. No solid In this way, hundreds of assembly important exterior dies are made inmodeler can hold the detail and problems are found before any metal is
problems are found before any metal is house. They are just beginning to use manipulate the data efficiently, they cut, comprising the majority of all such new nonlinear deformation software say. However, the wireframe models problems. There isn't enough detail in recently developed at the machine tool are useless when communicating with the CAD models to permit the rest to division. Up to now many problems the factory people or the tooling be found. Many of these involve flexi- had to be solved during die tryout. designers. So there are problems. ble items like pipes and wires. It is “Dedicated technicians are a big
The master data or the design ver- typical that thousands of design changes Japanese advantage,” said Kobayashi. sion of them are also passed to super- to improve assembleability will be made Stamping plant people visit NTC often computers for various analyses (see after prototypes are built.
during early styling design because (see laboratory tour below). Converting data There is some indication that Nissan above) the master data are frozen early. for use by these programs takes a long is beginning to adopt modular assem- While some of this visiting may be done time, another problem that adds time, bly, similar to what shipbuilders did by passing CAD files around the neteffort, or number of people required. 20 years ago. Examples given were entire work, most of it is done face to face.
CADII data are also used by indus- dashboards built up in advance and A big problem, other than die design trial engineers for doing manufactur- then inserted into the car with a robot. errors, is die design changes forced by ability and assembleability analyses. The This is, in fact, not new, since (1) you other design changes. These cause a assembly path of a part (we saw a wind- can't install efficiently things in a dash- stream of changes throughout the final shield washer tank) can be visualized as board once it is in the car and (2) GM year before launch. Thus the environthe engineer moves a view of it around and VW have used robots to put in ment of change Kobayashi referred to the screen using knobs. He can see dashboards for several years. However, above makes the use of overlapped interferences, view from several angles, the module approach has a bright future, jobs less of a strain than might be try different paths, and so on. There is although it may require some redesign supposed. no analytical support for this. He can of the vehicle's structure. Sticking one The use of CAD data to find probalso attach images of tools to the heads part after another into the engine lems and do designs prior to making of screws or nuts to evaluate access for compartment is getting harder and any prototypes obviously saves time tools. These issues are of the most harder to do. However, more success at and money. However, it appears that importance in designing the engine getting around this problem with this process is most successful during compartment, which is the most crowded modules will mean that maintenance the three prototypes that NTC builds. area. The success of this process depends and repair will become harder and When the plants start getting involved entirely on the experience of the harder! The modules are likely to be in prototypes, the use of CAD data engineer.
big and heavy and require special tools wanes because (1) the prototypes are While there is no analytical support and hoists to remove later.
much more compelling, (2) they are for assembly analysis, there are criteria.
easier for people who did not do the original design to see and understand, • Complex multicolor 3D wireframe engineers design things that only senior and (3) those CAD tools and data are rendition of the engine compart- engineers could before. One case is NTC's, not the factory's. This is a typi- ment. This picture occupies about design of trunk, brake, and axle systems. cal problem of “ownership!” The only 20 to 30 MB. They anticipate need- Another nonartificial intelligence data easily shared by all are the exterior ing 40 or more MB. Yoshida claimed application is use of 3D modeling to shell descriptions.
that commercial CAD packages study how to package parts in pallets
cannot manipulate this much data. and shipping containers and to see how Tour of CAE Facility
I argued this point with Kobayashi to use last year's jigs and fixtures for
later, giving examples from ship- this year's parts. CAE facilities are used to do exte- building that I was familiar with. Considerable effort has been made rior styling and many complex analy- But he countered that they discuss to export this design software system to ses, plus actual design engineering. The this problem with SDRC, who is a Nissan's suppliers. Nissan and IBM have facilities comprise two Craysupercom- joint venture partner with them, and a joint venture that supports the software puters, two IBM 3090s, two Unisys 2200s
and sells it. A low cost workstation plus several more large mainframes for
version is also for sale to small suppliers. supporting 3D CAD drawings, solid • View out the front of the car from At many suppliers there are now modeling for casting and forging design, the driver's viewpoint. This view was hundreds of terminals running CADII and routine data management. The not constructed from the master data on both host type computers and Yokosuka engine design facility has its file but from elsewhere, maybe from engineering workstations. Kobayashi own Unisys 2200 for designing castings scratch. In the future all these pro- said that Toyota and Unisys have a and forgings using solid models. The grams will be linked with common similar joint venture for selling Toyota's IBM and Unisys computers support
CAD to its suppliers. Problems arise 1,400 graphics terminals. Many of these
trying to sell to a large supplier like terminals use SDRC's I-DEAS pack- • Crash test simulation. This was a Hitachi, which has its own CAD software age. The mainframes also support replay of Cray output. The software and hardware, and to overseas suppliers I-DEAS in IBM's form (CAEDS) but for this is called PAMCRASH, who have different data standards. U.S. all agree that performance is poor and purchased in France and modified suppliers have a data format standard, workstation implementations are bet
for example, that forced Yoshida to ter. In addition, there are over 3,000
write a data translator. PCs of various types, including 1,000 • Aerodynamics simulation. This is Macs. The penetration ratio for PCs is also Cray data being replayed. They History and Goals of about 0.5 (one PC for two engineers). can analyze wind drag but not noise. Nissan's CAD Efforts The computers at NTC are networked This was a surprise to me. I have worldwide with similar computers in seen a paper presented that described CAD tool development and user the United Kingdom, the United States, Nissan efforts to calculate noise support, including support of business Spain, and Australia. The same soft- generated between the engine data systems, is provided by 220 people ware is supported at all sites.
compartment and the ground. This in the R&D Systems Department. In Several demonstrations of CAE were simulation shown to us on this visit 1985 it had only 70 people. given:
explicitly represented airflow under CAD at Nissan has three main goals the car, however.
and several historical threads: • Very realistic rendering of car exte
rior shape, including reflection of • Plastic flow simulation in an injec- 1. Data processing, especially test data light off the body. This is a surfaced tion mold. This is a typical commerwireframe model. It becomes the cial software capability.
2. Design specification control and master for all later design.
parts data trees
I asked if noise from engines and
showing dashboard, seats, and shift puter and was told yes.
troublesome problems, or to help junior