rick@cs.arizona.edu (Rick Schlichting) (05/21/91)
[Dr. David Kahaner is a numerical analyst visiting Japan for two-years under the auspices of the Office of Naval Research-Asia (ONR/Asia). The following is the professional opinion of David Kahaner and in no way has the blessing of the US Government or any agency of it. All information is dated and of limited life time. This disclaimer should be noted on ANY attribution.] [Copies of previous reports written by Kahaner can be obtained from host cs.arizona.edu using anonymous FTP.] To: Distribution From: David K. Kahaner ONR Asia [kahaner@xroads.cc.u-tokyo.ac.jp] Re: Sumitomo Metals use of AI and Expert Systems 18 May 1991 ABSTRACT. Some examples of AI/Expert systems at Sumitomo. INTRODUCTION In a recent report on Nippon Steel (nippon.st, 18 March 1991) I described several applications of computers at Japan's largest steel company. Lest readers infer that this kind of technology is limited only to the largest, Sumitomo Metal Industries sent me a copy of their February 1991 technical journal (unfortunately in Japanese) describing their applications of AI and Expert systems in steelmaking and related activities. Sumitomo is a large company, but as a steelmaker, it is not even in the top ten. I had already reported on an expert system tool containing more than 300 rules that they have applied to a large steel wire finishing plant (flexible, 12 Dec 1990), and this is item #8 below; the others report new activities. In that report I also explained that Sumitomo was entering the workstation business and developing high reliability RISC based workstations which they could sell with ready to go applications. Article 2. below is a specific example of this approach. The quoted text below is from the abstracts of the papers. A common thread on all the applications is to "make know-how common". That is, move expertise from the experts to a wider base within the steelmaking environment. Contents of Sumitomo Metals, Vol 43#1, 1991. 1. Applied Technology of Artificial Intelligence and its History for Sumitomo Metal's Applications. "Sumitomo Metal Industries, Ltd has developed various expert systems for controlling the abnormal stages of a plant, passing down design technologies and know-how to successors and so on which are good at these conditions in comparison with conventional computer systems. Here we first show the history of our activities for researching, developing our own tools, and expanding the applications. Next we describe the merits and problems of applying the expert system through experiences from ten or more sets of actual expert system building." In a speech at the 1990 ICOT Symposium, Tokai University Professor H. Karatsu made the following comments. They seem to me to be relevant here. "..115 new types of Japanese facsimiles were released in 1988. In other words, a new model ... every three days... [When] making a new model...slightly move the button on the right side nearer the left, to change the size of the package a little, or to change the color, and so forth. If you look at such modifications, each modification seems very trifling. However, after continuing such improvements for a few years, the machine turns into a completely different one. This is what Japan's technological innovation is... Most Americans think technological innovation is like hitting a home run in a baseball game. American batters are always trying to hit a home run, but often miss the ball, while Japanese batters are trying to make a base hit and a squeeze play. In the end, the Japanese team often wins the game.... Technology means, I think, to make something useful for man by applying natural principles that our predecessors found in nature... This process is completely different from science, where new discoveries and laws of nature are sought after. Science needs pioneering and foresight." 2. Developing of an Expert System Building Tool: SMI/MARKS-II. MARKS stand for Multi-Access-Realtime-Knowledge-System. The paper describes its design philosophy and illustrates some user interfaces. It also shows how the system can can link with systems such as on Sun, DEC, NEC, and IBM. As far as I can tell virtually all the applications below use this system. 3. Melting Instructions Expert System. "The engineering staff in the steelmaking plant use both technical knowledge and production know-how to write melting instructions. The instructions are written to inform the operators of the target chemical composition and target temperature, and to give important details of special related conditions. To improve the efficiency of the technical staff and ensure successful transfer of technology, we have systematized this procedure by applying the technology of an expert system. This system is a "design-type" expert system which runs on a large host computer which works in conjunction with conventional systems. This paper gives an outline of the system and discusses the factors involved in designing a practical expert system." 4. An Expert System for the Material Design of Large Diameter Steel Pipes. "An expert system for the material design of large diameter steel pipes has been developed for (1) passing down design technologies and know-how to successors, (2) improving design quality, and (3) reducing designing time. Given a customer's production requirements on dimensions, strength properties and toughness properties, it produces optimal production specifications such as chemical ingredients and rolling and heat treatment conditions in consideration of production costs, stability of products' quality, etc. In is an experimental rule-based system and is composed of a knowledge base, which contains the know-how of material design specialists, a production result data library which is referred to as a base data for designing, and a newly developed hill climbing optimization logic which is used for adjusting values of factors in order to improve contradicting interrelated material properties. It is implemented as man-machine conversational system and is easy to use. Even a beginner can design a material in 10 to 30 minutes." 5. A Material Arrangement System for a Hot Strip Mill. "The material planning system of a hot strip mill consists of a material arrangement system and a casting and rolling planning system. The former system's function is to assign orders to stocked slabs and the latter's is to request for slab supply to the proper steelmaking process. A fully automated assignment system of slabs, which is the kernel function of a material arrangement system, has been developed by using an expert system. Technology succession and improvement of work efficiency have been realized by extracting complex rules from the slab assignment job and by organizing them into the knowledge base. This expert system can also prevent a chance loss of an assignment, which contributes to a reduction in the number of stocked slabs." 6. Expert System for Blast Furnace Operation. "Sumitomo Metals has once succeeded in computer control of blast furnace operation under a normal furnace state by use of a mathematical prediction model. Recently, for the establishment of a completely automatic and intelligent control system, a hybrid expert system composed of both a math model and experimental rules was developed as an application for artificial intelligence. In the abnormal furnace state, control by experimental rules performed to achieve stable furnace operation for the prevention of a significant reduction in hot metal production. Further, diagnosis rules are provided in the system to determine the cause of abnormality and to select necessary actions for long term control. The system was introduced into the Kashima No. 1 blast furnace in October 1988. The hybrid expert system is successfully applied in actual operation." 7. A Real-time Expert System Applied to the Mold Bath Level Control of a Continuous Caster. " Fluctuations in the mold bath level of the continuous caster greatly influence the surface quality of cast slabs or billets. It is difficult to maintain an optimal bath level by a PID controlled and a slide gate vibration controlled with fixed parameters, because the characteristics of the casting condition fluctuate during operation. To solve this problem, an expert system method has been applied to the mold bath level control system. In this expert system, operator know-how regarding control parameter adjustment is represented in the form of a knowledge base and that knowledge base is driven by an inference engine when a significant fluctuations in the mold bath level occurs. This system has been applied to a continuous round caster, and it has effectively regulated mold bath level fluctuations by adjusting the controlling parameters to optimize the state of operation when the fluctuation occurred." 8. Coil Transfer Control Based on Expert System. "Sumitomo Metals has been promoting the adoption of automatic control systems, aiming for efficiency of productivity and material flow. Recently, we applied knowledge engineering to the field of automatic control systems and developed a hot coil transfer scheduling and control system based on AI for the hot cold finishing line at the Kashima Steel works. This system started operation in September 1988 and has operated very smoothly since then." 9. An AI System for the Facilities Monitoring On-line System. "In 1988 a diagnosis support expert system for equipment was applied to the facilities monitoring on-line system (FAMOS) at the Wakayama Steel works of Sumitomo Metals. This system, which used the MARKS-II, produced by SMI, is an on-line AI system. The characteristics of the system are as follows. (1) Can be operated together with all FAMOS's terminals joined with all the optical fiber networks around the works. (2) Can infer with measured data by FAMOS. (3) Can get diagnosis results for equipment using a knowledge data base that is structured with expert knowledge and experience. (4) Needs to infer in short periods so the operator can get exact information in real-time. This AI system is expected to work as an effective maintenance support tool." 10. Fault Localization Expert System for Hydraulic Screw-down of Tandem Cold Mill ant Kashima Steel Works. "The hydraulic screw-down of a cold strip mill is a complex device consisting of a mill system, a hydraulic system, and a control system. When trouble occurs in the hydraulic screw-down, it needs expert operators, expert mechanical-maintenancemen and expert electric- maintenancemen to discover the cause. In addition, this requires too much time. We have developed and employed the following two new systems. The Conditions Monitor is to prevent troubles. The Fault Monitor is to discover the cause of the trouble quickly and easily even for non- experts." 11. An Expert System for Automated Computer Operation. "A rule-based expert system for automated computer operation was developed using know-how from a procedure-based system. The system watches console messages and controls mainframe computers and their networks by commands or replies. The expression of system status and production rule is formalized. This information is stored in a relational database. The problem is fit for production systems. Thus, we can develop rules using half the man-power of the procedure-based system. Automated computer operation does not only make the operation more stable, but also makes know-how of the operation common." -----------------END OF REPORT-------------------------------------------