Site: Toshiba Corporation
Kawasaki-shi 210, Japan
Date Visited: March 24, 1992
Hosts: Dr. Seiichi Nishijima
Dr. Ken-ichi Mori
Toshiba's 1990 reported net sales were $22.9 billion. The number of employees totalled 72,000 (these data do not include subsidiary groups). Information/ communication systems and electronic devices comprise 57 percent of sales. R&D expenditures were 7.9 percent of sales. Between 1989 and 1990 Toshiba's sales in information and communication systems went from 28 percent to 57 percent. The corporation is therefore paying at lot of attention to computer-related products and technologies. We visited the Systems & Software Engineering Laboratory (SSEL), founded in 1987, and one of several corporate laboratories within the Toshiba R&D organization. SSEL has three research departments and a technology transfer department. The three research departments are: (1) AI and human interfaces, (2) distributed, fuzzy, and neural net systems, and (3) software engineering. The three divisions use the results of the research of other departments; for example, fuzzy logic is a part of the next generation ES tool.
We were given an overview of Toshiba's AI activities by Mr. Nishijima, followed by more detailed presentations: Mr. Kohno talked about Toshiba's ES tools; Mr. Tanaka discussed a design system that uses case-based reasoning; Mr. Kohno presented a model-based expert system for flexible plant control, a joint project with ICOT.
Toshiba's responses were the most extensive of any organization's.
Approximately 500 expert systems have been developed for both internal and external use, of which about 10 percent are in routine operation. Design and Planning/Scheduling are the major growth application areas. Within design, the principal tasks are large scale integration (semiconductors) and printed circuit board design.
The most successful expert system is a paper production scheduling system for the Tomakomai Mill of Ohji Paper Co., Ltd. The system uses 25 kinds of pulp, which are combined in 10 paper making machines to produce 200 different paper products. There are hundreds of constraints to be satisfied. The system employs a top-down hierarchical scheduling strategy, starting with scheduling product groups, then individual products, and then line balancing. The time required to produce a monthly schedule was reduced from three days to two hours.
Toshiba also reported data on a microwave circuit design system, called FIRE, built with an internally developed tool called Debut. FIRE captures the design process for highly parametric design problems. The system runs on Toshiba's AS4000 workstation, is C-based, and interfaces with microwave circuit simulators and a mechanical CAD system. The primary benefits of the system are speed-up of problem solving and accumulation of design knowledge.
A representative fault diagnosis system has been developed for Kyushu Electric Company, and is in routine use by Kyushu operators. The system diagnoses faults and restores operation to an electric power system. The fault diagnosis system has 900 rules; the fault restoration system has 600 rules. The system was built using TDES-3, a real-time application shell that uses rules and frames for knowledge representation. The development team consisted mostly of Toshiba employees, with domain experts supplied by Kyushu.
Our hosts also reported on a diagnostic system for a subway station facility, called SMART-7, which was built for the Tokyo Eidan 7th line. The system was built with a diagnostic knowledge acquisition support tool called DiKAST. SMART-7 is implemented as a support module that detects malfunctions in the air conditioning facilities. The system contains 1,600 frames, and runs on a AS4000 workstation. It was built by three system engineers in three months.
Another expert system is used for placing electronic components on printed circuit boards. The knowledge base consists of about 70 rules and 8,500 functions, and was built on top of the ASIREX tool. The ES is integrated with a PCB CAD tool called BOARDMATE, a commercial product developed by Toshiba. The system took three years to develop, with an estimated labor cost of three person-years. The system has resulted in a factor of ten speed-up in problem solving.
A small knowledge system (110 rules, 32,000 lines of C code) that Toshiba sells is MARKETS-I, a decision support system to determine the suitability of opening a convenience store at a particular site. Estimation accuracy is improved with the use of this system.
ESCORT is a banking operations advisor system that is used at Mitsui Bank. It plans the most appropriate procedure to get the computer banking system back in operation following an accident. The system has about 250 rules and 900 frames, and was built using a LISP-based expert system shell called EXPEARLS. The GUI was written in C. The system runs on the AS3000 workstation.
In the area of software engineering, Toshiba has developed an automatic programming system for sequence control. This system generates a control program for a steel plant from high-level specifications. It analyzes and refines the spec, generates code, and retrieves program modules. This is a fairly large system: 2,900 frames, 320 rules, and a library of 190 program modules. It was written in LISP, using an internally developed frame-based knowledge representation language with object oriented facilities. Twenty person-years went into its development, over a four year span. The system has resulted in cost reduction and an improvement in the quality of the sequence control program designs. Test and verification are performed manually.
Toshiba's systems do not now use multiple sources of expertise, but the company is trying to do so in its newer systems. Many ESs are implemented with a combination of a shell/tool plus a programming language such as C or LISP. Toshiba has several training courses, ranging from a one-day basic course, to a two- to three-week application development course, to a multi-week advanced topics course. About 10 percent of research funds go into training. An important element of Toshiba's methodology is to use task-specific shells, such as PROKAST or DiKAST.
ESs selected for implementation are chosen by a systems engineer or researcher. This technology is used only when conventional data processing doesn't work. The pre-specified selection criteria are performance and practical value. Toshiba also looks for an economic justification. Usually the same people are used in all phases of application selection, development, insertion into the operational activity, maintenance and redesign.
Toshiba's largest project to date is a 5,000 rule system for diagnosis and control of an electrical power generator.
Our hosts characterized their progress as steady.
Toshiba has developed several sophisticated shells for particular types of problems. These are oriented toward enabling the end-user organization to develop its own detailed knowledge bases.
Toshiba's first generation shell is MYEXPERT, a rule-based tool that runs on J-3100 personal computer and is commercially available.
Toshiba's second generation shells include:
Toshiba's domain shells (developed on ASIREX) include:
Another tool is KASE (Knowledge Acquisition Support Environment), which supports knowledge acquisition so that ESs can be developed by domain experts without the involvement of knowledge engineers. KASE has been used for diagnosing defects in color picture tubes and in casting parts, for support of repairing hard disk drives, and for fault diagnosis of a traffic signal system.
With respect to advanced KBS activity, Toshiba is seriously exploring model-based and case-based reasoning, machine learning, natural language understanding, and advanced assumption-based truth maintenance systems (ATMSs). We received twelve technical papers from our hosts on a wide variety of subjects and systems, all in English.
Toshiba's pipeline of R&D through applications seems well balanced.
A new development, which uses recursive case-based reasoning (to two levels), deserves further study (Tanaka 1992). The idea is to use other cases, rather than rules, to modify a case that doesn't quite fit.
The model-based plant control system designed to handle unforeseen problems was particularly interesting. A conventional, heuristic, shallow reasoning layer handles anticipated problems. It defers to the deeper, model-based layer when unable to handle a problem. The deep reasoning layer has a fuzzy, qualitative physics simulator.
Toshiba is the only organization that is paying much attention to a formal problem solving framework into which the heuristics fit. Others are mostly following a straight heuristic "interview the expert" approach.
Although several software development types of applications were counted in the statistics, we did not see any examples. It is possible that, at Toshiba, software engineering research is focused upon only the newer types of software.
In general, Toshiba appears to be very advanced in the application of knowledge-based systems. The range of application types (problem solving situations) is quite broad. The proportion of applications reaching operational status is high.
The range of techniques and technologies being integrated at Toshiba is quite broad (fuzzy expert systems, CBR with neural nets, simulators, conventional order processing systems, etc.). Toshiba has clearly moved beyond the introductory stage in that few if any of their applications are stand-alone systems.
Toshiba's strategy: train engineers and provide them with tools.
The JTEC team was impressed by the use of "soft" justifications for systems efforts. Toshiba does not seem to be encumbered by the ROI/MBA mentality that pervades U.S. thinking. Continuous process or product quality improvement, for example, is sufficient justification.
Our hosts view was that the most important features in the next generation tools are: task specific tools, knowledge acquisition tools, cooperative problem solving, and advanced reasoning, such as case-based and qualitative reasoning.
In general, Toshiba has organized its research activities into five to ten year projects which are conducted in corporate research labs, 3-5 year projects in the development labs, and the more immediate problems in the business group. (This type of organization seems to be common -- for example, at Mitsubishi.) For software the path is more direct. The Systems and Software Engineering Lab (SSEL), although under the corporate lab for organizational purposes, does research and basic development (such as prototyping of tools) and supports the work of operational groups. It also does corporate planning for system and software technologies.
Although a relative newcomer, Toshiba seemed the most balanced and best organized of the computer manufacturers we visited (NEC, Fujitsu, Hitachi) with respect to their approach to ES technology development and the ES business.
Toshiba's second generation shells, recently introduced, run on SPARC workstations. The company's ASIREX shell costs 4 million yen, and its workstations run from 1.8 million yen to 10 million yen in the maximum memory configuration.
Toshiba's ES business is steady, but below expectations. The number of people using ES is increasing. Toshiba's strategy is to have its software engineers study AI technology and use it internally.
Toshiba gave very little business data. The company reports a 17 percent hit rate on getting from start of an ES project to deployment. Applications are moving to planning from diagnostics. According to our hosts, LISP machines are finished. Toshiba appears to have many good internal applications.