Using an intelligent supervisory system to simulate automatic operation in a crushing plant

An intelligent supervisor using expert system tools and applied to a crushing plant has been developed with a structure based on a model of a human operator. It consists of a knowledge base containing a qualitative model of the process and a rules base with the heuristic logic used by an operator when supervising a plant. The simulation results obtained show the potential of this type of system to supervise the automatic operation of a plant. There has been very significant development in artificial intelligence over the last two decades. Within this development, two approaches can be distinguished, (Winston, 1984). The "model approach," in which the main objective is to understand and model the be¬havior of living systems. The second approach is the "engineering approach." Here, systems capable of performing tasks usually requiring human intelligence are created. As part of this latter approach, knowledge engineering and its main achievement, expert systems, have been developed. Expert systems arose from the need to incorporate the abilities of a human expert in solving realistic problems in which the conventional approach has great difficulties. The first successful applications of expert systems were in medical diagnosis, geophysics, and fault detection. However, since 1971, when Fu (1971; defined intelligent control as the inter action of artificial intelligence and automatic control, some significant developments in the use of artificial intelligence tools for process control have been achieved (Astrom, 1986; Pinto 1988; Fahrenkrog, 1988; and Porter 1987). The modern control theory has made important advances. And, although successful in some applications, it has failed to cope with the practicalities of many industrial processes. As a result, the automatic operation of a plant has been confined to a small region around a normal operating point and to periods without any great disturbance. One of the main reasons for this is the lack of accurate and structured knowledge of the processes involved. Despite this, operators applying their knowledge, experience, and a little intuition, often obtain good results controlling the plant in such situations. Considering the above, the next obvious step is to include similar reasoning capacities in a computational system. The development of such an expert supervisory system prototype applied to a mineral crushing plant is presented. Problem description The problem is to conceive a control system with the ability to handle quali-