(A) Regenerative expert system approach for the maintenance of a large-scale expert systems

Abstract

In recent years, expert systems have proven to be efficient in solving a wide variety of real-world problems. As many industries have made widespread use of diverse expert systems, expert systems become a large- scale and complex one. Furthermore, the nature of expert system-separation of control knowledge and domain knowledge-makes it difficult to develop and maintain expert systems. Maintaining expert systems is critically important but very difficult like most conventional software systems. The focus of most expert system maintenance has so far been concentrated on the maintenance of knowledge bases. However, the maintenance of many industrial expert systems such as scheduling expert systems incorporates not only the knowledge base but also the inference engines and program modules. Therefore, the maintenance of expert systems calls for the development of software and maintenance methodologies such as the software reuse. The approach we have taken to handle this situation is the regenerative expert system approach which can modify the system to accommodate the changed situation starting with a standard initial system. The detailedness level of modification we have attempted is the design specification of rule statements, program modules(let us regard the inference engines as a special type of module set),and module parameters. We have proposed the regenerative expert system approach as a general framework of developing and maintaining a large-scale expert systems. The approach is illustrated with the case of assembly line scheduling expert system for shipbuilding. REGENESYS, a prototype system which helps the development of standard expert system and its modification according to the changed specifications, is developed using expert system development tool UNIK.