SHORT-TERM PREDICTION OF SILICON CONTENT IN PIG-IRON

Prediction and control of the silicon content of pig iron is an important but difficult task in the operation of the blast furnace process. A model for time-series modeling of the silicon content in small- or medium-sized blast furnaces is presented. The model makes use of recent measurement information from the furnace and considers the dynamics of the furnace hearth in a simplified way. Based on ladle-wise chemical analyses, which provide information about the short-term behavior of the silicon content, the parameters of an autoregressive vector model with irregularly observed data are estimated. A practically continuous estimate of the silicon content is obtained, which may act as an indicator for the operator, who has to decide about possible control actions to be taken. The model also predicts the evolution of the silicon content during the next few hours.