Hybrid predictive modeling and simulation of the Teniente converter

Abstract

The principal copper pyrometallurgical technology employed in Chile is the Teniente converter (TC). The smelting and converting processes involve complex dynamics with strong interactions among manipulated variables, state variables, and product characteristics. Furthermore, the converter products are periodically extracted in a semi-continuous process. Previous applications of model predictive control in TC operation have shown good performance. The recent development of an online-sensor for measuring the levels of the liquid phases opens up new possibilities for control of the process. The TC prediction model must accommodate discrete states for the white metal and slag tap holes. When the TC is inclined for maintenance, the injection system is disconnected and slag evacuation may occur via the converter mouth, thus affecting liquid level dynamics. The three possible incline positions or angles determine three discrete states, each one corresponding to a different set of liquid dynamics. Describing these situations therefore requires a hybrid dynamic model. This paper presents a hybrid grey box model for the Teniente converter process that combines continuous and discrete variables to describe six different operating mode.