Knowledge-based control of rotary kiln

The paper describes an expert system for real-time control of rotary cement kiln operation. Rotary kiln is the central and the most complex component of cement production process. Its task is to take an appropriate mixture of input material (“raw meal”), and to gradually burn and bake it to produce clinker, coarse-grained pieces of cement which are then transported away from the kiln and milled in a special mill to get the cement dust. The kiln itself is a complex cylindrical device, consuming fuel to get preheated to a high temperature necessary to produce clinker. It rotates around its axis and the raw meal dust sticks adhesively to its walls, thus getting burned and baked. To control the kiln operation successfully, a number of measurements must be performed continuously in real-time, and a number of control signals (i.e. setpoint variables) must be generated simultaneously. It is not always possible to use classical control algorithms to handle this problem appropriately. Many input variables change asynchronously over time and may generate disturbance in the normal operation. It requires substantial knowledge and experience from the operator in order to keep the kiln operation smooth enough. Therefore, we developed a real-time expert system to perform the task of kiln operation control. The system is based on a previously developed methodology for building real-time expert systems. The paper overviews the methodology, describes the problem domain in detail, and discusses some important design decisions and implementation details of our expert system