A DESIGN OF ITERATIVE LEARNING CONTROL-SYSTEM AND ITS APPLICATION TO A BATCH POLYMERIZATION REACTOR

The conventional PID method is said to be inadequate for the temperature control of batchwise polymerizations with rapid heat generation of reaction at temperatures somewhat higher than the initiation temperature of the reaction. Overshooting phenomena of reaction temperature can be practically avodied by a forced switching of heating to cooling at the so-called turning-point temperature, which is set according to past experience of operation. It is troublesome to determine the turning point by trial and error for every case where polymerization reactor conditions change dynamically or with time. We have developed a new iterative control system in which a hypothetical reference pattern to the above reactor's jacket temperature is determined automatically from the viewpoint of system similarity. The new system enables us to approximate batchwise polymerization reactor dynamics by the ARX model, using observed data at the first trial operation, and also to set a hypothetical reference pattern for the subsequent operation. The hypothtical reference pattern can be adjusted by iterative use of the error-modifying function. This new control system was applied to the batchwise exothermic reaction of PPS to confirm the design philosophy. It indicates that the new system is usable for the temperature control of large-scale polymerization reactors. © 1990, The Society of Chemical Engineers, Japan. All rights reserved.