For all their sophistication, distributed control systems (DCSs) have produced mixed results when measured by bottom-line yardsticks such as product quantity, quality, and economics. Sometimes, a good operator relying on personal experience with the process can make it perform better by setting the controls to semiautomatic mode and manually adjusting setpoints. In such instances, artificial intelligence techniques can use an operator's experience to good advantage. Expert systems can capture the process operations knowledge of a mill's best operators to adjust process control setpoints directly or offer advice on adjusting them. Expert systems are easier to use than traditional programming techniques because they use natural language and do not require line numbers in the knowledge base. The contents of the knowledge base are usually expressed in easily understood "if-then" statements; the lack of line numbers means that the contents can be arranged in any order and still function perfectly. This article explains how selecting an appropriate expert system shell and spending the time to ask questions of good operators makes it possible to develop successful supervisory control strategies. These can be applied to processes in lime kilns, blast furnaces, power plants, and various chemical plants.
