Professor Myung-Hwan Kim Vice President Korea Academy of Industrial Technology 70-6 Yangjae-Dong Seocho-gu, Seoul, Korea Tel: 575-3955 Fax: 577-5488 Unfortunately, Professor Kim was not in his laboratory when I arrived, but several other faculty and graduate students showed the system to me. KAICUBE is more or less a standard hypercube. KAICUBE I uses a 68020 CPU and 68882 floating point unit. Maximum communication speed is 120M bps. At this time a 64-node KAICUBE II has been built, with peak performance of about 64 MFLOPS. The first machine, with 8 nodes, was completed last year. The group has just begun to build a 128-node machine (KAICUBE III) with the new, faster Intel 1860 processor. They estimate that it will be completed at the end of 1992 and will have 5 GFLOP peak performance. It will also have improved communication, up to 320M bps. they want to learn how to do it themselves so that they don't have to buy the technology from outside. As far as I can tell this is much the same motivation that drives the POSTECH project. This is an excellent way to train students and develop basic research technology. Its most important use is internal, but the Japanese have shown that this approach can be successful at rapidly catching up with research in more advanced countries. What is not clear, though, is whether, once rough parity has been achieved, breakthrough ideas will also be forthcoming. KOREA SCIENCE COUNSELOR I also stopped in for a courtesy visit with Mr. Kenneth Dewitt Cohen Counselor for Scientific and Technological Affairs U.S. Embassy 82 Sejong-ro, Jongro-ku Tel: 732-2601, x4210 Mr. Cohen emphasized that it was very useful for U.S. scientists who were planning to visit Korea to make his office aware of their plans. He has a good overall grasp of scientific activities in the country and can often make excellent suggestions about additional worthwhile travel. The KAICUBE group briefly showed me some of their work on the architecture and software development of the machine, which they have done essentially all themselves. As far as I can tell there are no documents available in English. The little documentation I did see described various applications, such as solving Laplace's equation, sorting, fast Fourier transform (FFT), simulated annealing, and database query acceleration. But I had no opportunity POSTSCRIPT to discuss this with any of the actual users. I was shown the results of one benchmark (traveling salesman problem) in which the KAICUBE II was 2.5 times slower than the Cray 25. Nevertheless, as this is in almost every way a traditional hypercube, I asked them specifically why build it when they could buy one much more easily. Their answers were frank and illuminating. Basically, After I wrote my first draft of this report, Sang K. Cha [chask@eclipse. stanford.edu] pointed out to me that the KAICUBE research project recently moved from KAIST to the Korea Academy of Industrial Technology (KAIT). KAIT was started this year by the Ministry of Commerce and Industry and is led by a Seoul National University (SNU) professor in international economics. He also felt that SNU and KAIST were Korea's two top universities and that I should be sure to visit them on my next trip. Cha also suggests that during my next visit I should see those organizations that are associated with the Ministry of Communication (MOC). He pointed out that unlike the Ministry of Science and Technology (MIST), MOC controls the telephone company fcalled the Korean Telecommunication Agency (KTA)] and directs some of its revenue for research in technology and policy. For instance, while the Electronic Telecommunication Research Institute (ETRI), the biggest research institution in Korea, is officially under the supervision of MIST, it is financially under the control of MOC. In addition to KTA, the Data Communication Corp. of Korea operates computer communication networks and leads government computerization projects with ETRI. MOC also established a research institute on communication technology policy and a few other smaller scale research and supporting organizations. SUMMARY Korea is far behind Japan and the United States in most computing research but has a very pragmatic view of what they want to accomplish. Essentially, this is to have their own computing infrastructure even if this requires some reinventing of known ideas and technologies. Many senior Korean scientists have been trained or spent substantial time in the West but have returned because of opportunities that did not exist in their own country until recently. Thus they are keenly aware of what is going on outside Korea. At the moment there are few Western scientists (of non-Korean origin) spending sabbatical or other research time there. Relations between Korean and Japanese scientists are cordial, but Koreans have a strong affinity to collaborate with U.S. and European scientists. David K. Kahaner joined the staff of the Office of Naval Research Far East as a specialist in scientific computing in November 1989. He obtained his Ph.D. in applied mathematics from Stevens Institute of Technology in 1968. From 1978 until 1989 Dr. Kahaner was a group leader in the Center for Computing and Applied Mathematics at the National Institute of Standards and Technology, formerly the National Bureau of Standards. He was responsible for scientific software development on both large and small computers. From 1968 until 1979 he was in the Computing Division at Los Alamos National Laboratory. Dr. Kahaner is the author of two books and more than 50 research papers. He also edits a column on scientific applications of computers for the Society of Industrial and Applied Mathematics. His major research interests are in the development of algorithms and associated software. His programs for solution of differential equations, evaluation of integrals, random numbers, and others are used worldwide in many scientific computing laboratories. Dr. Kahaner's electronic mail address is: kahaner@xroads.cc.utokyo.ac.jp. NIPPON STEEL KIMITSU WORKS A visit to Nippon Steel's Kimitsu Works is described. by David K. Kahaner huge, with nearly half a million NIPPON STEEL The Kimitsu plant is due south of Tokyo (about 75 minutes by fast train), across Tokyo Bay in Chiba prefecture, and facing across the bay toward the industrialized cities of Yokohama and Kawasaki. It occupies 10 million square meters of flat, partially reclaimed land. There are about 13,000 people at the facility; less than half are Nippon Steel employees, and the others are contractors. In fact, Nippon employment at the site has been decreasing since the mid 1970s. Steel production at the site was mildly decreasing until 1988, at which time it took a large jump due to consolidation of other facilities. The fact that the company produces more steel at Kimitsu now than it did at its peak in the 1970s, but with about 15% fewer people, suggests that their automation procedures have been working. Employee attrition has meant that hiring has not been a problem, and that the plant has ample applicants for the few hundred positions that it has each year. (Mandatory retirement, except for the most senior positions, is at 60 years of age.) On the other hand, we were also told that applicants are not as good as 14,200 they once were, presumably because more Japanese students are going to college, and hence the average ability level is going down. There are almost no Ph.D. hires. Nippon/Kimitsu, like many Japanese companies, prefers to hire younger scientists and train them in the specific skills needed in their organization. By company, steel production (in thousands of tons) in 1989 was as follows: Japan consumes about 80% of its production, approximately half in the construction industry, 17% for automobiles, and 8% for industrial machinery. Although only a small amount of crude steel is imported to Japan, almost 100% of the iron ore and coal, the major raw materials used in steel making, is imported, primarily usx, U.S. from Australia, Brazil, and India. By comparison with other parts of Japanese industry, the steel industry is not exceptionally healthy. Nevertheless, a company like Nippon Steel is Nippon Steel, Japan NKK, Japan Thyssen, W. Germany Bethlehem Steel, U.S. 28,000 12,900 12,300 11,900 11,400 11,000 11,000 Kimitsu is the largest of Nippon Steel's plants, producing almost 10 million tons of steel each year. Nevertheless, the key to making it economically successful is to use enough automation so that much smaller customized orders can be economically produced. This was emphasized to me repeatedly during my visit. For example, each slab from the furnace can be associated with a different order and has individual specifications and destination. This philosophy is consistent with some of my earlier reports (see the Scientific Information Brief on robots on page 4) about Japanese industry. No steel plant is really clean. This one was probably much better than most. There are many new buildings, but some were clearly part of the original plant, now 25 years old. A modern research and development (R&D) center (1989) is within the plant site and undertakes research closely related to the production line (this was not part of our visit). In addition, just outside the present plant, three large R&D laboratories that are currently in other parts of Japan will be brought together into a brand new Research and Engineering Center of about one million square meters with 1,200 scientists. Topics to be studied include artificial intelligence (AI), fuzzy logic, robotics, and virtual reality. The Kimitsu plant sends about 100 engineers and scientists overseas each year, and about 1,000 trainees per year from various countries are invited to Kimitsu. With respect to international cooperation, for example from the United States, we were told that very few basic scientists are invited, and all of these are from industry. I'm no expert on any aspect of steel making. However, with me on this visit were Dr. Iqbal Ahmad (Army Research Office, Tokyo), a materials specialist who understands steel technology, and three representatives of the commercial section of the U.S. Embassy, who were able to explain many of the economic details. The Nippon Steel host was Mr. Mutsumi Ohji Tel: +81-439-52-4111 Our visit was coordinated by Dr. R. Yamaguchi U.S. and Foreign Commercial Service U.S. Embassy 1-10-5 Akasaka Minato-ku, Tokyo 107, Japan Tel: +81-3-3224-5058, -5060 the second year of its second cycle.) Molten pig iron is tapped eight times each day and then the hot metal is transferred to special rail cars for delivery to the oxygen furnace, which is in another building. Today's newest plants move these two closer together, even into the same facility. At Kimitsu, once the molten metal gets to the oxygen furnace, the remainder of the casting operation is continuous, with the steel moving through one of the longest buildings that I have ever seen. Red hot slabs go in one end, and coils, sheets, and other things come out the other, all still so hot they can't be touched. The entire facility has remarkably few people in attendance and appears to be thoroughly automated. Slabs weighing up to 45 tons each can be processed. These can be formed into very long coils, or left as plates that can be individually formed to any length between 3 and 25 meters. The basic steel making process is KIMITSU-COMPUTING simple in principle and exquisitely complex in execution and detail. Iron ore, coal, and limestone are dumped in layers into a furnace (blast furnace) and melted. The molten "pig iron" goes to another furnace (basic oxygen furnace) where oxygen is injected in order to burn out and adjust the carbon content. Molten steel then goes to be cast (formed) into long rectangular slabs or long blooms (square cross section). Slabs eventually become flat products such as sheets, coils, pipes, or tubes, while blooms will become flanged beams, bars, and other sections. The trick is to do this quickly, efficiently, and safely, while at the same time maintaining consistency and quality in products with many different chemical and structural properties. The blast furnace (one of three) we saw at Kimitsu has an internal volume of over 5,000 m3 and runs continuously for about 10 years, after which it is shut down and repaired. (It is currently in My own interest in the blast furnace was associated with some AI techniques that are used to control it. After 25 years of steel making experience at Kimitsu, about 1,200 "If-then" rules have been developed (knowledge base) and integrated into an expert system, called ALIS, that monitors and adjusts various parameters associated with the furnace. The real time inputs to this are from 1,000 sensors. The system was written in-house, in C, and has a completely parallel maintenance system, so that new rules can be added or old ones modified by the operational staff. I think that this is fairly unusual in such a large system. Of course, I was not shown any of the actual code, which is proprietary, only some displays pictorially representing the blast furnace and its operation. In the control room I asked about fuzzy reasoning. The interesting aspect here was that the blast furnace supervisor was very well aware of what |