Development of a knowledge-based hierarchical control structure for process automation

Fuzzy logic is applicable in representation and processing of knowledge in some types of knowledge-based control. The technique is particularly useful when the plant that is to be controlled is complex, incompletely known, and difficult to model either analytically or experimentally, but when a knowledge base is available in the form of if—then rules containing fuzzy descriptors. The standard practice of applying fuzzy logic in control systems is to replace a conventional direct controller with a rulebase and an inference mechanism that is based on fuzzy set theory. Thus, in the standard fuzzy logic control, the knowledge-based controller is located in the low-level control loop itself. In the present research, a significantly different approach to standard fuzzy logic con trol, one that is particularly useful in process automation, is considered. The knowledge-based control system developed in this research has a hierarchical architecture, where knowledge-based decision making that depends on fuzzy logic, is employed for high-level functions like process monitoring, tuning, and supervisory control, leaving the low-level direct control to conventional controllers. It is argued that since fuzzy logic is primarily a method of artificial intelligence, the proper place for such a tool would be the upper levels of a hierarchy rather than in low-level direct control, where the fastest and most high-resolution data processing take place. A general model for a hierarchical fuzzy system is introduced, which uses transitional and combinational operators. Some characteristics of these operators are explored. Hi- erarchical fuzzy systems are shown to be characterized by several heuristic features such as information resolution, fuzziness of information, and the required data processing intelligence. Some preliminary relationships between these parameters are explored. It is argued that fish processing is one application where the knowledge-based hier- archical control system that is developed in this research, is appropriate. The rationale for this choice is given. As the application testbed of the developed technology, an auto mated workcell for fish processing that has been developed in the Industrial Automation Laboratory is employed. An on-line system is implemented for monitoring and tuning of the workcell, which incorporates computer vision, knowledge based tuning, servomotor operation, and conveyor control. The attractiveness of employing fuzzy logic in the con text of a data processing hierarchy is illustrated, by means of a case study of application, where large quantities of low-level information that is generated by various sensors are abstracted through the use of fuzzy-logic based processing. The resulting information that has a lower resolution but more amenable to knowledge-based decision making, per mits one to perform more intelligent data processing at a reduced computational burden, and by making use of available experience and expertise on the particular process plant.Applied Science, Faculty ofMechanical Engineering, Department ofGraduat