« iepriekšējāTurpināt »
STATE OF THE ART IN JAPANESE
COMPUTER-AIDED DESIGN METHODOLOGIES FOR MECHANICAL PRODUCTS: REPORTS ON INDIVIDUAL VISITS TO COMPANIES AND UNIVERSITIES
The author spent 3 months in Japan as a temporary liaison scientist with
the Office of Naval Research Asian Office to survey Japanese use of
the first issue of 1992; this is an appendix that contains detailed
by Daniel E. Whitney
UNIVERSITY OF TOKYO: NEW students. Tuition is high but appar- leads in shipbuilding, this department
structure. When it existed, it dealt priENGINEERING
The typical course ofstudy is 4 years marily with engines and other ship
for a bachelor's degree, 2 more for a machinery, not with ship structure or 5 June 1991
master's, and 3 more for a Ph.D. The other traditional naval architecture.
undergraduate curriculum is almost all Elsewhere in Todai there is now a Background
classroom courses while graduate study Department of Shipbuilding and Naval
is mostly laboratory work and thesis Architecture. Prof. H. Inoue is in the Mechanical with only a few classes. This is impor- Three years ago the ME DepartEngineering Department and is cur- tant to understand in view of the sub- ment decided that it was losing sturently head of the new Mechano- ject of this report, which depends heavily dents or would soon, with the defectors Informatics Department. How this new on curriculum reform.
going into more modern technologies department came into being is the subject
based on computers and information of this report. New Departments
sciences. (Inoue said this twice during The University of Tokyo (Todai) is
our talk.) The response was to "restrucJapan's most prestigious. Its graduates Mechanical engineering used to be ture” and modernize the curriculum. go into the best government and uni- split into three subdepartments called The pressure to restructure came versity positions, including most of the Mechanical Engineering (ME), not only from trends visible in student new hires into Todai's own faculty. The Mechanical Engineering for Produc- registrations but also in general from buildings are quite old, solid reinforced tion, and Marine Engineering. The latter the rush of technological change in concrete, and hard to modernize. Bud- came into being about 20 to 30 years society and industry. Japan identified gets are tight. Almost all the students ago as Japan became a prime ship- information-intensive products as straare self-supporting, including graduate building country. Since Japan no longer tegically important as early as 1970 with the launching of the PIPS (Pat. Dept of Mechanical and Industrial Broad ME includes industrial engitern Information Processing Systems) Engineering (Broad ME)
neering and production engineering. 5-year national project and has pur
Both design and computing appear in sued this area intensely since. Obvi- 1. Production Systems, Manufactur- all three subdivisions. Students majorously mechatronic products will prolif- ing Systems
ing in any one of these three take courses erate and engineers will be needed to
from the various chairs, with 50% recomdesign them. Industry is quite inter- 2. Machine Creation and mended from the home department ested in this new restructuring.
and 25% each from the other two. There Inoue noted that the restructuring
are no required subjects. The requirebegan 3 years ago and the Ministry of 3. Systems Engineering, Security ments for what we call "humanities” Education took until this January to Engineering
subjects, basic science, and math are give final approval. Todai is a national
satisfied in the first 1.5 years when the university subject to the Ministry's 4. Design Engineering
students attend a different campus. This governance. I do not know if there is an
way of setting up the curriculum may equivalent of ABET other than the 5. Human Systems Engineering have been adopted in order to reduce Ministry, but I doubt it. He also notes
conflict between the advocates of the that all the debate, curriculum crea- 6. Industrial Systems, Transportation new curriculum and those of the old tion, and course design occurred dur- Systems
who usually ask in such debates what ing this time, so the big fights are over
mechanical engineering really is. The and the new structure is fully in effect. 7. Humanware Systems Engineering new structure actually moots this ques
(donated chair by JR East Japan) tion in a very realistic way, acknowlDepartment and Curricular
edging the fact that the old curricular Structure Dept of Mechano-Informatics (New ME) and discipline boundaries have long
since been destroyed by external events The new department structure recog- 1. Electronics and Computer in and it is necessary to build new ones. nizes "traditional deep" ME, broad ME, Machinery (digital systems, micro- According to Inoue, the purpose of and mechano-informatics (new ME). computer, interface, micro-machine) the Department of Mechano
Informatics is to enhance the research Dept of Mechanical Engineering (Deep 2. Mechanism and Control (mecha- and education of computer-intensive ME)
tronics, control theory, mechanics mechanical engineering. Primary and mechanisms)
research fields include: 1. Strength of Materials and Structure
3. Pattern Information Processing • creation of intelligent machinery such 2. Fluid Dynamics
(sensors, signal processing, image as robotics and mechatronics
processing, visualization) 3. Thermodynamics
• computer-intensive design and 4. Software Engineering (algorithm analysis of mechanical systems 4. Mechanical Vibration and design, programming languages, Mechanics
• introducing new functions or
approaches into machinery such as 5. Material Physics and Tribology 5. Computation Mechanical Engineer- bionic functions, neuroscience, and
ing (computational mechanics, micro-machines 6. Energy Conversion, Combustion simulation, FEM analysis, CAE) Physics
• advanced human-machine interface, 6. Bio-Mechanical Engineering (bio- virtual reality, cognitive engineer7. Heat and Mass Transfer
mechanics, neuro engineering, cog- ing, etc.
nitive engineering) 8. Mechanical Science, Measurement
Onsubsequent visits I hope to delve Instrumentation
7. Information Systems Engineering deeper into such questions as the rela
(robotics, artificial intelligence, tion between university training and information systems)
company training and whether the university thinks any one student can out before trickling it down to the Tomizuka also said that curricular really learn all the things that are offered. undergraduate curriculum. This can take reforms like those described above What should a competent design engi- many years and lacks a department- actually happen fairly frequently in Japan neer know in a world of mechatronics? wide strategic approach. It also lacks a and do not represent the revolution Since there are no required subjects, methodology for removing outdated that is implied by my report. only “strongly recommended” ones, the material, leading to crowding in the The comments of Tomizuka and department has not taken a rigid stand undergraduate syllabus. At Todai the Ono raise a difficult question: if Japanese on these points. [This topic is addressed graduate curriculum has so few classes university education in engineering is in the report on Hitachi Construction that this method may not be available. so shallow, how come Japanese prodMachinery.)
The methodology at Todai is not uct engineering is so good? The answer I raised the question of the place of totally clear to me, except that the apparently lies in the additional educaalgorithms in this curriculum. It may pressure came from within the depart- tion the youngengineers get on the job, seem odd to relate algorithms to mechan- ment, apparently, and not from the plus such factors as lifetime employical design but Inoue agreed imme- dean. The methodology for selecting ment, extensive use of past design data diately that this is an essential ingredient. elements of the new curriculum is also on new designs, and the length of time Many complex products are algorithm- not clear, except that the chairs focus an engineer keeps the same job driven by their embedded micropro- on areas that are related to their research. responsibilities. cessors. Many have complex user inter- This creates expertise but does not An advantage of this kind of educafaces and multiple internal states, both guarantee that generic material will be tion is that it sets the pattern for “unimechanical and electronic. Thus a sense taught or that the students will obtain
taught or that the students will obtain versal experience,” meaning that an of algorithms is essential for a compre- a balanced education.
engineer more easily learns and prachensive design approach.
What is clear is that the change was tices many fields during his/her career. A related question is why algorithm- quite radicaland has defined “mechan
quite radicaland has defined “mechan. “Mechanical" engineering graduates do aware students don't go into computer ical engineering” in a way that would not feel a professional commitment to science (CS). The simple answer is that be almost unrecognizable at many mechanical engineering but rather to there is no CS department in Todai's schools in the United States.
their employer, who may alter their engineering school! There is a CS
professional concentration as a result department in the School of Science, Postscript
of assignments and training. These however. I did not learn much about
alterations apparently do not cause much what it teaches. The electrical engi- This report was distributed in draft discomfort. Cross-trained engineers neering (EE) department in the School form to many U.S. educators and drew perhaps can understand each other's of Engineering deals mostly with power an interesting response from Prof. design problems, making concurrent and information systems, including Masayoshi Tomizuka, Vice Department engineering easier to implement. signal processing and vision. Most U.S. Chairman of Mechanical Engineering U.S. engineering students, on the universities have CS departments or at UC Berkeley. Tomizuka did his Ph.D. other hand, devote a lot of their educaCS divisions of EE departments. At research with me at MIT in the early tion to becoming “mechanical engiTodai such competition does not exist, 1970s after getting his SB and SM from
1970s after getting his SB and SM from neers,” for example, and might feel leaving a clear path to ME for such Keio University in Tokyo.
their school time was wasted if their students who also have a mechanical Tomizuka said (and Dr. Kozo Ono employer tried retraining them as EEs. bent.
of Hitachi Construction Machinery Similarly, U.S. companies expect new
Company and a Todai graduate con- hires to function productively soon Discussion
firmed) that Japanese undergraduate after being hired, just because of the
engineering education is broad and focussed character of their education, Many universities in the United shallow. One is exposed to many fields and would not think of retraining them States have trouble changing their but learns almost no deep knowledge. to a different field. Engineers thus rapidly curricula radically, in spite of obvious Tomizuka said he was rised by the become specialized and less able to reasons to do so. At the Massachusetts depth of the MIT doctoral qualifying communicate with engineers in other Institute of Technology (MIT) I saw exams and had to study very hard to fields. leading professors introduce new mate- learn the material to pass them.
Therefore, education, career paths, rial at the graduate level and prove it
and company training (or lack of it) are symbiotic in both countries.
PROF. FUJIMOTO, DEPT. OF and no esthetics at all. (For more about in isolation since numerous other fea-
which may not be captured in the ques
tionnaire. 7 June 1991 Research Methodology
Fujimoto says that companies are
willing to participate in such studies Background
Clark and Fujimoto pursue a style because their own data are disguised
of research that is often called “deter- and they are able to benchmark themProf. Takahiro Fujimoto is a recent mining best practice.” It differs markedly selves against regional averages. graduate of Harvard Business School, from accepted academic research in Sometimes they can decode the data a where he collaborated with Prof. Kim most fields and at most business schools, little and discern information about Clark. They have already written a book where scholarly research is typically single companies, but such opportunities and are continuing their joint studies. about economic models, financial analy- are limited. The subject is management methods in sis methods, or investment strategies. the automobile business world-wide, This method is practiced by conducting Importance of This Kind with a prime focus on the product field studies, sending out questionnaires of Research development cycle; specifically, the (typically 80 to 100 questions), and doing issues studied are lead time, develop- some statistical analyses on the results. It is essential that different methment productivity, and total product Thus it is more like anthropology than ods of managing the design and proquality. The questions they addressed management science. The results often duction process be identified and
make fascinating readingand are taken compared. It is well known that differ
seriously by forward-thinking people ent companies in the same industry can 1. How do auto companies in Europe, in the subject industries. However, the differ in productivity by as much as a
Japan, and the United States organize studies often require a leap from "form” factor of two. Increasing attention is the product development process? to "content" in the sense that the char- being given to the length of the product
acteristics of a company that can be development cycle in particular because 2. What are the main bottlenecks that gleaned from questionnaires and
and it gives so much competitive advantage cause the process to take a long interviews may not be the root causes to companies. They can follow the time?
of the differences between company market as it changes, absorb new techachievements.
nology into their products sooner, and 3. What regional differences are there For example, a recent MIT study build experience in their technologies in the length of the process? (Ref 1) revealed that car factories with and in the design process itself at a
more democratic management styles, higher rate. The ability to “climb the 4. What managerial techniques and more flexibility in job classifications, learning curve" of design and technol
organizational practices account and more automation were able to make ogy faster has been cited by Gomory for the relative differences found more cars in more model varieties with and Schmitt (Ref 2) as a major factor in between companies and regions? fewer defects than factories that lacked national and international competition
any one or two of the above character- based on national productivity and Their past studies have focussed on istics. Yet it is not obvious that if one economic strength. body and chassis design of cars and the started up a new factory and included Two distinct approaches to speeddesign of the manufacturing systems all these characteristics one would ing and improving the product develthat make those parts. Although engines automatically obtain high quality, high opment process have been identified. generally take longer to design than model-mix production capability. The These may be called roughly the bodies, bodies change more frequently, authors of the study show statistical management-intensive approach and so their development time, which is significance in their survey results but the technology-intensive approach. The long as well, tends to dominate the this study method does not really lend former emphasizes management and design cycle for individual car models. itself to the typical methods of statisti- organizational methods while the latFuture studies will focus on engines, cal analysis. For example, there is no ter emphasizes use of computer-aided which have relatively more engineer- way to create a control set; you cannot design and similar techniques. Coming analysis and less (though NOT no) create a double blind environment; you panies using managerial methods esthetics behind them, and semicon- cannot establish a controlled set of employ design teams from different ductors, which have even more analysis features whose effects are to be tested technical disciplines. Those pursuing the technical track make use of com- developed. Fujimoto believes that this phase activities are finalized. Intense puters for design or for internal com- is the right sequence for obtaining the communication is used to minimize munication. Others use modern best processes, at least in the auto the effect of such risks, as discussed computer-controlled manufacturing industry. Others feel that the technical below (Ref 3). equipment to obtain better uniformity challenges of complex products are so of output, hence higher quality. No great that without help from computers, Importance of Overlapping company uses one technique alone. no company can surmount them merely Design Functions and The differences are matters of degree with management techniques. Interfunction Communication only.
Therefore the book is still open on In some industries it is taken for what is the best approach or mix of The overlapping design process is granted that design technology is essen- approaches to improving product devel- often called “Concurrent Design” (CD) tial. Modern complex products require opment methods. Clark and Fujimoto (Ref 4). Other synonyms are Concurso much data to describe their design, have learned a great deal about the rent Engineering and Simultaneous so many calculations to determine auto industry, and a summary of their Engineering. All refer to a process during performance, and so much attention to findings follows.
which product designers and producdetail during fabrication and assembly
tion system designers exchange inforthat human capabilities are severely Fujimoto/Clark Major Findings mation in order to maximize the prochallenged. In the semiconductor indus
ducibility of the product. In more sophistry, human capabilities in design and Fujimoto and Clark approached their ticated CD environments, factors such design checking were far surpassed over study as a management problem and as marketability and field repairability 15 years ago; only use of large scale asked management questions. They did are also taken carefully into account. computing permits modern micropro- not, in fact, investigate use of comput- The contrasting situation, which occurs cessors to be designed at all.
ers, either in extent or kind of use. The in the defense industry as well as in American companies in particular implications of this are discussed below. high performance companies such as are known for using computers heavily At the same time, Fujimoto knows that Mercedes, is that product performance in many aspects of design and produc- CAD/CAM has had a major effect on dominates design, with all other faction whereas Japanese companies are car development time, referring to tors being secondary or not considered not so well known for this. In fact, some Toyota's finding that wide introduc- during design proper. They are considof the most productive Japanese com- tion of CAD terminals reduced lead ered later, during production system panies, such as Toyota (the Just in Time time half a year.
design, or not at all. This is called (JIT) method for making cars) and Their basic finding is that Japanese “throwing the design over the wall.” Ishikawajima-Harima Heavy Industries car companies on average take a year The trouble with allowing perfor(IHI) (the moduler method for making less to design a car than U.S. or European mance to dominate design is that the ships), achieved their famous produc- companies and that this shorter devel- other factors cannot be served except tion efficiencies without large scale use opment time results from two factors. by including them in the original design of computers for factory management. The largest one is overlapping of phases or by changing the design later. But Since Japanese performance in these of the process that normally are accom- changing the design is prohibitively industries sets the standard for the world, plished sequentially. The second factor expensive once it is complete. A design it is somewhat ironic to find, at least is that some of the phases are accom- often represents a long chain of interbased on current outside knowledge, plished faster, although some of the dependent decisions and roads not taken. that computers do not play a leading time saved is used to perfect the design Changes that are more than cosmetic role.
rather than to shorten the overall pro- may threaten to unravel the entire chain. The purpose of my own study in cess further. The main method by which Producibility or repairability problems Japan has been to pursue the question overlapping is achieved, according to that are discovered after design are of how computers are used in the design Fujimoto/Clark, is via intense commu- said to cost 10 times as much to fix as if process, not the factory management nication between product designers and
nication between product designers and they were discovered when the design process. It appears, however, that production tooling designers. This was only drawings; if discovered during Japanese companies have spent the last communication allows the tooling production the cost is 100 times as 10 to 20 years refining both design and designers to critique the design and to much. In addition, poor producibility manufacturing processes independently begin key phases of their own work. and repairability lead to low quality of computers and are now applying This process can be risky because sec- and a bad image for the product in the computers to the methods already ond phase activities start before first marketplace.