A novel structure of real-time expert control system for process industry

A new structure of on-line real-time expert control system for process industry is proposed in this paper. The overall system can be hierarchically visualized as three levels. The highest level performs the supervisory functions, such as overall system performance monitoring, possible system components failure detections and diagnoses; and commands generations regarding to the necessary controller parameter tunings or structure variations. It is in this level that the expert control system stores the knowledge and makes the decisions. The second level utilizes some sophisticated signal processing algorithms, such as adaptive filtering, system identification, state estimation, to pre-process the raw measurement data coming from the sensors. So that the procedure of decision-making and structure of the knowledge base become simpler and more concise. It is this level that makes the present structure unique in comparison with the other existing expert control systems. The lowest level consists of the conventional control hardwares which execute the control commands. A prototype system is under construction on an IBM PC-AT microprocessor using PROLOG for knowledge manipulation, decision-making and FORTRAN for numerical computations.