Ferramenta preditiva para o nível de líquidos em cadinhos de altos-fornos

Abstract

This dissertation presents a system for monitoring the level of liquid in the hearth of the blast furnace by monitoring the electromotive force, or EMF, generated in the blast furnace shell. Also shown are the procedures for installation of sensors to capture the signal, the signal processing in order to make it stationary, independent of the thermo-chemical state of the hearth and highly correlated with the level. Further, it is performed through the modeling of the signal time-series using Box & Jenkins methodology. Finally, using the model found, it was used to predict the behavior of the signal 30 minutes ahead. The main contribution of this dissertation is to propose a strategy for analysis, identification, filtering and estimation of EMF signals through ARIMA models resulting on a accurate forecast for 30 minutes of the liquid level inside hearth.