Particle Swarm Optimization Based Tuning of a Modified Smith Predictor for Mold Level Control in Continuous Casting

Mold level variations are a serious productivity and quality problem in continuous casting process. This work proposes a level control structure based on the Aström's modified Smith predictor able to improve the reduction of the bulging effect. Unlike conventional methods, this control strategy decouples the disturbance rejection from the setpoint response and therefore can be independently optimized. Within this scheme, the bulging rejection specifications are formulated as a H∞ problem and the tuning parameters are designed through a particle swarm optimization approach. Simulation results confirm that the proposed architecture is more effective than the ones currently implemented in real plants.