Adaptive Control of Meniscus Velocity in Continuous Caster based on NARX Neural Network Model

Abstract

Meniscus velocity in continuous casting is critical in determining the quality of the steel. Due to the complex nature of the various interacting phenomena in the process, designing model-based controllers can prove to be a challenge. In this paper a NARX neural network model is trained to describe the complex relationship between the applied current to an Electromagnetic Brake (EMBr) and the measured meniscus velocity. The data for the model is obtained using a laboratory scale continuous casting plant. Adaptive Model Predictive Control (MPC) was used to deal with the non-linearity of the model by adapting the prediction model to the different operating conditions. The controller uses the EMBr as an actuator to keep the meniscus velocity within the optimum range, and reject disturbances that occur during the casting process such as changing the casting speed.