Fuji Electric has developed and shipped various types of intelligent systems making use of AI, fuzzy theory, and neural networks.
The aim of intelligent systems is to promote advanced factory automation and efficient development of software.
This paper describes the features and trends of these systems and introduces Fuji Electric's current status, research and development, problems and plans for the future.

AI technology has been playing a more important part in the automation and labor saving of plant operation under increasing expectations. The optimum operational solution to total business systems is demanded particularly in these days and various approaches by application of AI technology have been attempted. Under the circumstances, AI application methods reliable and easy to use and near-perfect products are earnestly desired. This paper introduces Fuji Electric's attitude to the application of AI technology to products and the lineup of its products.

There are many jobs in power system operation which need operators, experience and professional knowledge. Fault location is one of these jobs. The development of a system to support fault location was intensely hoped for.
This paper describes a system with a real-time operating function installed in a control center of Shikoku Electric Power Co., Inc. In this system, relay functions are represented by a protection zone and time setting, and protective simulation for reasoning is adopted. This method provides quick and precise location, reason for estimation easy to understand, and flexible adaptation to system expansion.

AI technology has long been introduced in water and sewage plants. Fuzzy control of chemicals dosing control is a pilot example. This paper outlines AI applications in those plants and introduces system configurations and application examples. An example in waterworks is an operation program for a large water distribution system including a mathematical planning method in a practical form. An example in sewage treatment is rainwater pumping plant control that selects a suitable operation out of efficient operation to meet rainfall and operation giving priority to safety. Both systems are operating satisfactorily in the actual plants.

The automation of overhead-traveling cranes for handling loads in steelworks and other factories requires transferring loads precisely to the destination and without giving a swing. Fuji Electric has developed a fuzzy control system that coordinates operating systems for positioning and swing prevention. The result of simulations and actual operation tests using this control system proved that distances to destinations and swing angles converge after about 2.3 times the swing period of loads when positioning and swing prevention control is effected.

This paper introduces a power system operation supporting system applied to the Nippon Kokan (NKK) Keihin Works Energy Management Center. The system applies an expect system based on experienced operators' knowledge to enhance the security of electric power transmission in the large steel plant and to reduce unexperienced operators' routine load in the Electric-Instrumentation-Computer-unified environment. The system has functions for proper commissioning in various system configurations such as system restoration support in case of blackout, switching operation guidance, and simulation of a fault and protective relay operations.

Fuji Electric has developed supporting software AIMAX-P for scheduling a factory operation plan in the field of factory automation. This tool is characterized by preparing knowledge expressions that dispense with the description of standard models of production equipment and allocation rules of orders to the equipment. The standard models for expression are classified into three models of factory equipment (equipment models), products (product models), and production orders (operation models). This paper describes the detailed expressions and how AIMAX-P is applied to the field of factory automation.

The defrosting control of open showcases is an important control on which goods freshness control performance and running cost are dependent. Fuji Electric has developed a fuzzy defrosting controller that aims at advanced freshness control and energy saving by applying fuzzy control to defrosting control. The fuzzy defrosting controller precisely grasps surrounding conditions and frosting amount and keeps the number of defrosting operations at a necessary minimum to attain the above aim. This paper describes an outline of the system and its effects citing the field test results.

The AI system is expected to support user's decision work and be able to update knowledge bases to meet changes caused by change in equipment easily. AI engineers, called knowledge engineers, often have to acquire empirical knowledge through interviews with skilled operators and other means. They strongly feel the technical difficulty as a bottleneck in knowledge acquisition. To make the work easy, Fuji Electric has developed knowledge acquisition AI software CHISHIKI, which makes knowledge input possible with a computer on a job site. The system has been used for the construction of our knowledge bases satisfactorily.

This paper describes an outline of neural network technology and applications by Fuji Electric. The neural networks include a multilayer model superior in pattern recognition and a mutually combined model superior in optimization.
As examples of the former, a system to distinguish automobile models by their outlines and a demand forecast system based on the past record of power consumption are introduced. As an example of the latter, a system to support the generation planning of a power system is cited. The examples show that neural network technology is effective and applicable to actual problems.

Fuji Electric has developed a new on-line fault analysis system with expert system and neural networks. The system considers the information on relay and circuit breaker operation, and fault currents and voltages obtained from advanced telecommunication systems, and makes a judgment about the fault point, fault type, and whether the fault is single or multiple. The system is constructed with FREXS/PS, that is a domain specific shell for power systems developed by Fuji Electric and can be easily installed in the control center as an on-line function.

The die department of Toyota Motor Corp. wanted improvement of estimation because the empirical formula used when the production plan had been made could not meet changes in production process conditions and estimated production time considerably differed from the actual one. The application of neural networks capable of meeting these changes by learning function was studied at study level with a good result. On the basis of the result, Toyota Motor Corp., Toyota Central R&D Laboratories, Inc. and Fuji Electric have developed an estimation system for die production time. This paper gives the outline, functions, and the evaluation of operation.