A method of system identification and adaptive control on regulation with stochastic disturbance and observation noise-an application to mold meniscus level regulation on a continuous steel caster

The authors describe a new adaptive control technique applied to a mold meniscus-level regulation system with stochastic disturbance and observation noise in a continuous steel caster. Essentially, a spectrum model based on a system transfer function is fitted to a sampled spectrum by using the maximum likelihood estimation method. This method has several superior features for practical use. First, it can provide parameter estimates without any arbitrary test signals such as a step input, so that the process is not disturbed. Second, owing to an FFT card, the signal processing procedure can be easily implemented on a personal computer. Third, it is possible to bound the frequency range for the identification. In addition, the system is easy to tune up.<>