Identification of the Multivariable Dynamical System of a Blast Furnace by Means of the Extended Maximum Likelihood Method

Abstract

For the automation of a modern blast furnace a precise mathematical model for the dynamic behaviour should be available. A blast furnace represents a complex multi-variable dynamical system. Therefore, its complete identification is very difficult However, the identification of a reduced quality model of the important high temperature zone of the blast furnace seems to be reasonable. For the experimental identification an extended off-line version of the maximum likelihood method (ML-method) is applied for modelling the blast furnace as a multivariable system. Considering all possible disturbances in the operation of a blast furnace and all errors which may occur during the measuring time, the quality model estimated by the extended maximum likelihood method (EML-method) describes the real process behaviour with quite accurate quality.