| The advances in computer technology, information processing, and control system theory in the last decade allow much more sophisticated control systems with higher degrees of autonomy than previously possible. A control system with a high degree of autonomy is capable of maintaining the desired performance under significant system uncertainty and outside disturbances and can accommodate system failures over an extended period of time without external intervention. In most cases, and especially for complex process systems, the benefits of these advanced control systems can far outweigh the additional controller complexity. Control systems with higher degrees of autonomy can provide increased overall process system performance due to better optimization of system interactions, less downtime because of better fault accommodation, and increased safety by allowing more human operator response time to anomalous events.; A high autonomy, three level, hierarchical control system concept is presented that incorporates new algorithms made possible by the recent advances in technology. Algorithm development and implementation for the middle supervision level of the control architecture is the focus of this research. Well-developed algorithms that address the fundamental implementation issues pertaining to supervisory reconfigurable control, anomalous event handling, and automatic controller design are the significant contributions of this research. Simple models of the Penn State Breazeale Reactor are used to demonstrate the implementation of the first two levels of the control architecture and to develop the new supervision level algorithms. The approach used in this research to achieve the desired system performance is to describe the outside disturbances and system faults as system uncertainties, and subsequently to analyze and partition these uncertainties. Modern robust control theory and fuzzy inference approximate reasoning techniques provide appropriate and compatible methods to produce desired system performance using these partitioned uncertainties. |
