Multitasking strategies in support of a knowledge-based operator companion

Today's operator, of nuclear-electric generating stations and chemical process plants functions in a data and information rich environment. In an effort to aid such operators, the structure of a real-time knowledge-based advisor is herein investigated for the central sampling system at Pt. Lepreau Generating Station. Human expertise (inherently symbolic or pattern recognition based) must be married to the speed and pervasiveness of computerized data acquisition and analysis which reside more in the domain of numeric processing. The schema advanced in this proposal achieves functional and temporal abstraction by using a tiered decomposition of the tasks, linked by real-time asynchronous agents acting through a blackboard. An on-line behaviour can be achieved by developing a specific instance in the small of functional and temporal abstraction for a real-time system using the anthropomorphic approach of a blackboard partitioned along the lines of manager - supervisor - technician. Any Overhead encountered in conforming an engineering domain problem to the suggested configuration is out-balanced by many gained advantages such as parallel pursuit of the primitive solution segments and the capability of changing focus of the system. DESQView, a multitasking DOS shell provided a test platform. A prototype system was developed to test message passing techniques and problem decomposability.